The Unz Review - Mobile
A Collection of Interesting, Important, and Controversial Perspectives Largely Excluded from the American Mainstream Media
 BlogviewJames Thompson Archive
Journey of 1.1 Million Miles
🔊 Listen RSS
Email This Page to Someone

 Remember My Information



=>

Bookmark Toggle AllToCAdd to LibraryRemove from Library • BShow CommentNext New CommentNext New Reply
Search Text Case Sensitive  Exact Words  Include Comments
List of Bookmarks

synapse labelled by lee

It is a great pleasure to see that a massive new study on intelligence has just been published, after years of work and also months of publication delays. Anything which can be done to speed up the publication of results is to be welcomed. Research has now moved to an international dimension, with disparate groups being managed and cajoled into cooperative ventures, a major undertaking that requires academia to develop new skills of diplomacy, coordination of disparate research groups, careful assembly of very different data formats and research protocols, and a sensitive understanding of individual egos and conflicting cultural and political sensitivities.

In contrast, all that is required of a commentator is patience, in this case a year of waiting.

Gene discovery and polygenic prediction from a genome-wide association study of educational attainment in 1.1 million individuals. James Lee et al. Nature Genetics, 2018

http://em.rdcu.be/wf/click?upn=lMZy1lernSJ7apc5DgYM8fw09BYcMPpe5594qF8ZKlg-3D_B-2Fvi2XVxdZyv3q46ooOzJWuNxuMdccQ4Qg-2Btclew6OCDyovz11vZ9mk5Z-2FJ8HP4lo3aKbMjlCB6Hl3RDKDMaH9Ilq1dKqilJ4aPLCsDEc-2ByjHcCRqVAwrvzgYElVjfqd6j4UdTccDzyeIM2ywQAQp6UH-2B3tMUVNN1AytbCx5h669-2BawS1tbA1m-2Fjf1iN5OcXklAgElh4fKNXtNp9GTtz7bHBBo3jCELyo0Vu9EIJZSbinLthTznebDTl0vpSk8E5

It is a very long paper, almost book size, because it has to cover so much ground. The table of contents lists 205 pages, of which 149 are explanatory text. Happily, supplementary figure 8 provides me with a picture I can use instead of having to explain everything. The picture (shown above) encapsulates the vast strides that have been taken in linking genetic research to the terra incognita of the synaptic gap which, if I thought about it at all as an undergraduate, seemed simply a mysterious chemical soup which passed message on upstairs to the brain, my main focus of interest. Now we have not only a diagram of the transmitter exchanges but linkage to the snippets of genetic code which lead to the actual processes.

First, a little comment about “years of education”. I think of this as a weak proxy for intelligence, and that it is used simply because the data are more widely available than intelligence test results. It will still be picking up interesting things, but will under-record pure intelligence. To look at the extent of this difference, look at other studies which have used actual intelligence test results, and you will find that they seem to tap into the same areas. When that is done, the correlation between intelligence and years of education is 0.7 which is perfectly respectable, and as high as correlations between Wechsler subtests. Also, see below for a comparison done within the study using two intelligence tests.

Second, in terms of the history of this research, this paper is EA3 (educational attainment 3rd sample) and follows on from previous work EA1 and EA2. Internationally there is some overlap and repetition in papers from other labs, and we probably need to explain these better. You may remember that last September I hoped “someone somewhere is keeping track of the overall picture, perhaps in a control room with multiple screens, like the NASA control centre of old, tracking the orbit of each SNP as it hoves into sight.” Not yet.

Third, the bulk of the paper is about the methods used to put together the samples into one big data base and all the corrections and assumptions which go into detecting the genetic signal within the noise. Some analyses, for example within-family studies, cannot be done reliably without 47,000 sibling pairs and they had only 22,135 identified, so they do only some restricted inferential work. A great deal has to be said about all these statistical matters, and the paper serves as a text for where the field has reached at the moment. Other leading labs will pile in with their detailed observations in due course. Please note the quaintly named “Winner’s Curse Adjustment”.

In this section the authors say:

Using a large sample of genotyped parent-child pairs from Iceland, the study documented that a polygenic score for EduYears constructed entirely from non-transmitted parental alleles predicts a respondent’s educational attainment. A plausible interpretation of this finding is that non-transmitted alleles are associated with EduYears through their effects on the child’s rearing environment. The effect of a polygenic score based on non-transmitted alleles was approximately 30% as large as the effect of a polygenic score based on transmitted alleles. An analogous analysis of height found that the effect of the non-transmitted-allele score was 6% as large as the effect of the transmitted allele score.

I find this difficult to take in, because the idea of the “child that you could have been” and “the parents you could have had” are new to me. However, the paper argues that an effect going from parent genotype, to parent phenotype, to offspring phenotype, is a very plausible explanation of the smaller effects inferred from the within-family studies than from the population GWAS. Assortative mating probably makes some contribution to the discrepancy as well.

As shown in Supplementary Table 38, in Add Health, a one-standard-deviation increase in the score is associated with a 4.7 percentage-point increase in the probability of completing high school incremental pseudo- R2=6.2%), a 15.6 percentage-point increase in the probability of completing college (incremental pseudo-R2=9.5%), and a 7.1 percentage-point reduction in the probability of having retaken a grade (incremental pseudo-R2=4.0%).

Here are the findings in a histogram:

Lee poly score and education achievement

Now consider, as Steve Hsu has done, the opportunity faced by parents who are having IVF because their genes contain a risk factor for Huntington’s chorea, or cystic fibrosis, or some other awful disorder, such that their petri dish embryos have to be screened before the best one is implanted in the womb. A doctor might say, very privately, to the parents “We have knocked out the genes for disease X as requested. All these 10 embryos will be fine. Would you like us to select the one of those 10 most likely to complete high school? Entirely up to you”.

At the request of a referee, Lee et al. have a go at using their polygenic score to predict the educational attainment of 1519 African Americans. It does not prove powerful, accounting for no more than 1.6% of the variance. This is a 85% come-down from the power of the score to predict European attainments, which they describe as an attenuation. However, this degree of attenuation is similar to that of 3 papers using European risk data to predict African American scores: 63% attenuation for education years, 88% attenuation for psychosis, and 85% attenuation for BMI. However, the predictive power of the polygenic score in other races is expected to decline purely from differing LD (linkage dysequilibrium) patterns; a SNP that tags a causal SNP in Europeans may not do so in Africans. The mere fact of a decline is not enough to say that the effects of the true causal sites differ in the two races. It may simply be that the SNPs point in a slightly wrong direction, but this is not resolvable at the moment. It would be good to have far more genetic and intelligence data on Africans, and then see how predictions based on them, our probable ancestral rootstock, predict European abilities. All that for later, when better data become available.

The authors see whether the polygenic score can predict actual intelligence test results: the Peabody Picture Vocabulary Test and the Henmon-Nelson Test.

The MTAG-CP score is more predictive than the GWAS-CP score, with an incremental R2 of 6.9% and again over the GWAS-CP score of1.8%.
In WLS, the MTAG-CP score is the most predictive of the four scores, with an incremental R2 of 9.7% and a gain over the GWAS-CP score of 2.7%.
In sum, the score can be used to predict the intelligence results to some degree, though once we get very large samples with intelligence test results then predictive power will be very likely to improve.

Here is a nice result, which validates work Heiner Rindermann did years ago, showing that it was better to have educated parents than rich parents.

Lee poly score and income

https://www.unz.com/jthompson/educated-parents-more-important-than

It barely needs saying, but the tissues these SNPs most enrich are in the brain.

This paper covers considerable ground, and does so in a careful way. I think it will have a massive impact. The sample size and the quality controls on the data will overcome doubts about the applicability of the results. That is, I assume that they will (some readers will wish to avert their eyes) and the excitement of this and similar papers will influence how we think about intelligence, education and heritability. It is part of an international effort to identify the biological causes of cognitive ability. This brings the whole project many steps closer to its goal.

 
• Category: Science • Tags: Genetics, IQ 
Hide 159 CommentsLeave a Comment
Commenters to Ignore...to FollowEndorsed Only
Trim Comments?
    []
  1. res says:

    It barely needs saying, but the tissues these SNPs most enrich are in the brain.

    And pituitary! (again) From page 121 of the Supplementary Material:

    When the tissues are ranked by heritability enrichment, the top result is frontal cortex (10.4-fold enrichment, P = 2.84×10–5), followed by pituitary (7.65-fold enrichment, P = 0.08) and broadly expressed (4.42-fold enrichment, P = 2.57×10–4). When the tissues are ranked by proportion of heritability, however, broadly expressed (proportion of h2 = 0.026, one-sided P = 6.25×10–7) overtakes frontal cortex (proportion of h2 = 0.019, one-sided P = 1.36×10–6). If we assume (unrealistically) no sampling covariance between these two estimates, then the difference between broadly expressed122 and frontal cortex in proportion of heritability is not significant (P = 0.30). The remaining results combined do not account for even half as much heritability as frontal cortex.

    Supplementary Figure 24 on page 198 shows this visually.

    Supplementary Figure 27 gives an interesting look at predictive power by chromosome. Chromosome 2 is both large and disproportionately predictive.

    P.S. Here is the caption for your lead figure:

    Supplementary Figure 8. Roles of Selected Newly Prioritized Genes in Neuronal Communication.
    The 59 genes listed in the figure were selected as follows. We began with the 30 gene-set clusters in Supplementary Figure 22 and dropped those that include gene sets that were implicated in a previous study of EduYears (Supplementary Table 4.5.1 of Okbay et al. 1). Of the 8 clusters that remained, we retained the 4 related to neuronal communication (DAG and IP3 signaling, associative learning, post NMDA receptor activation events, regulation of neurotransmitter levels).
    We identified the 460 DEPICTprioritized genes belonging to the exemplary gene sets representing these clusters (membership Z score > 2). Of these, the figure shows the 59 genes that appear in a figure or table of Fain352; these are genes whose functions are considered important for neuronal physiology

    Read More
    • Replies: @hyperbola
    This article has already been extensively discussed in two previous threads here at UNZ review. Now we have still more gobble-de-gook about a "paper" that needs 200 pages to hide the real conclusions.

    These “million person” studies are so under-sampled as to be nothing but random noise, which is why each one of them finds completely different sets of genes. Take 40 genes with 2 variants each (this is only a very small fraction of the genes that have been proposed to be correlated with things like “IQ”). The possible number of genetic variants is then 2 to the 40th power – that is more than ONE TRILLION combinations of those 40 genes with only two variants each. A sample size of one million persons corresponds to testing less than 1 part per million of those combinations.

    This leads to two inescapable conclusions.

    1. There are NOT enough people in the world to EVER do meaningful statistics on complex traits that involve >ca. 30 genes. ALL of these studies are measuring nothing more than noise artefacts.

    2. NO genetic test will ever be able to predict the “IQ” (or related things like “education achievement”) for an individual person.

    The question we should be asking ourselves is why we are being sold so much obvious snake oil from “psychologists”.
    ReplyAgree/Disagree/Etc.
    AgreeDisagreeLOLTroll
    These buttons register your public Agreement, Disagreement, Troll, or LOL with the selected comment. They are ONLY available to recent, frequent commenters who have saved their Name+Email using the 'Remember My Information' checkbox, and may also ONLY be used once per hour.
    Ignore Commenter Follow Commenter
    More... This Commenter This Thread Hide Thread Display All Comments
  2. Yes, the pituitary is trying to tell us something. Thanks for the more detailed list, which I skipped over.

    Read More
    • Replies: @Anonymous
    Now this is interesting, and I wonder if the pineal gland is involved too. As far as I know, these two do not have neurons but act by secreting hormones, which means their influence would appear to be indirect.

    This perked my interest because these two glands are traditionally held to be the physical correlations of the 6th and 7th chakras which are involved in attaining higher levels of consciousness. Higher consciousness grants clearer perception, which can be measured as higher intelligence.

    So then it seems we're not just talking about the correlation of physical structure and intelligence. Instead, we're talking about a three-way correlation: consciousness, intelligence, and physical structure.

    Is it allowed to bring consciousness into this, James Thompson? :-)
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  3. dearieme says:

    “It would be good to have far more genetic and intelligence data on Africans, and then see how predictions based on them, our probable ancestral rootstock, predict European abilities. All that for later, when better data become available.”

    Present-day Africans are not our ancestral rootstock. We are descended, the current wisdom teaches, from a small sample of the different sorts of humans who inhabited Africa at the time our ancestors left. Present-day Africans presumably share that descent while also being descended from ancient Africans from whom we are not descended. (In discussions like this I take it that “Africa” is shorthand for Sub-Saharan Africa.) Or so I understand – open to correction.

    Read More
    • Replies: @James Thompson
    Correct. Simply that non-Africans may have variants that Africans never picked up because they did not have to adapt to new biomes.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  4. @dearieme
    "It would be good to have far more genetic and intelligence data on Africans, and then see how predictions based on them, our probable ancestral rootstock, predict European abilities. All that for later, when better data become available."

    Present-day Africans are not our ancestral rootstock. We are descended, the current wisdom teaches, from a small sample of the different sorts of humans who inhabited Africa at the time our ancestors left. Present-day Africans presumably share that descent while also being descended from ancient Africans from whom we are not descended. (In discussions like this I take it that "Africa" is shorthand for Sub-Saharan Africa.) Or so I understand - open to correction.

    Correct. Simply that non-Africans may have variants that Africans never picked up because they did not have to adapt to new biomes.

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  5. Factorize says:

    Approximately 100 of the SNPs found were described as causal. Does causal truly mean causal? Would causal SNPs apply to all human populations? Yet, it is still possible that even causal SNPs might only be causal within certain environmental contexts.

    Read More
    • Replies: @James Thompson
    Why is it not fair to describe DNA as causal?
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  6. @Factorize
    Approximately 100 of the SNPs found were described as causal. Does causal truly mean causal? Would causal SNPs apply to all human populations? Yet, it is still possible that even causal SNPs might only be causal within certain environmental contexts.

    Why is it not fair to describe DNA as causal?

    Read More
    • Replies: @res
    He is probably thinking of the issue of detected SNPs being in LD with causal DNA (I think this could be SNPs OR structural variation) rather than causal themselves. The populations comment is because LD can differ between populations. Some of your Davide Piffer posts mention this either in the posts themselves or in the comments.

    Some relevant excerpts from the paper:

    The results also raise a number of possible targets for functional studies. Among SNPs within 50kb of lead SNPs, 127 of them are identified by the fine-mapping tool CAVIARBF24 as likely causal SNPs (posterior probability>0.9; Supplementary Table 10). Eight of these are non-synonymous, and one of these eight (rs61734410) is located in CACNA1H (Supplementary Fig. 9), which encodes the pore-forming subunit of a voltage-gated calcium channel that has been implicated in the trafficking of N-methyl-D-aspartate receptors25.
     

    We used the tool CAVIARBF24,48 in a fine-mapping exercise to identify candidate causal SNPs. We used the 74 baseline annotations employed by stratified LD score regression as well as 451 annotations from Pickrell47. We applied a MAF filter of 0.01 and a sample-size filter of 400,000 and only considered SNPs within a 50-kb radius of a lead SNP. We computed exact Bayes factors by averaging over prior variances of 0.01, 0.1 and 0.5; we set the sample size to the mean sample size of our considered SNPs; and we added 0.2 to the main diagonal of the LD matrix because we used a reference panel for LD estimation. To incorporate annotations, we used the elastic net setting with parameters selected via fivefold cross validation. The resulting annotation effect sizes and list of candidate causal SNPs are given in Supplementary Tables 37 and 10. Regional association plots of four noteworthy candidates are shown in Supplementary Fig. 9.
     
    The Supplementary Material has more discussion on pp. 119-121. Supplementary Table 10 has a list of the causal SNPs. One thing that surprises me is how few of them are coding SNPs.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  7. res says:
    @James Thompson
    Why is it not fair to describe DNA as causal?

    He is probably thinking of the issue of detected SNPs being in LD with causal DNA (I think this could be SNPs OR structural variation) rather than causal themselves. The populations comment is because LD can differ between populations. Some of your Davide Piffer posts mention this either in the posts themselves or in the comments.

    Some relevant excerpts from the paper:

    The results also raise a number of possible targets for functional studies. Among SNPs within 50kb of lead SNPs, 127 of them are identified by the fine-mapping tool CAVIARBF24 as likely causal SNPs (posterior probability>0.9; Supplementary Table 10). Eight of these are non-synonymous, and one of these eight (rs61734410) is located in CACNA1H (Supplementary Fig. 9), which encodes the pore-forming subunit of a voltage-gated calcium channel that has been implicated in the trafficking of N-methyl-D-aspartate receptors25.

    We used the tool CAVIARBF24,48 in a fine-mapping exercise to identify candidate causal SNPs. We used the 74 baseline annotations employed by stratified LD score regression as well as 451 annotations from Pickrell47. We applied a MAF filter of 0.01 and a sample-size filter of 400,000 and only considered SNPs within a 50-kb radius of a lead SNP. We computed exact Bayes factors by averaging over prior variances of 0.01, 0.1 and 0.5; we set the sample size to the mean sample size of our considered SNPs; and we added 0.2 to the main diagonal of the LD matrix because we used a reference panel for LD estimation. To incorporate annotations, we used the elastic net setting with parameters selected via fivefold cross validation. The resulting annotation effect sizes and list of candidate causal SNPs are given in Supplementary Tables 37 and 10. Regional association plots of four noteworthy candidates are shown in Supplementary Fig. 9.

    The Supplementary Material has more discussion on pp. 119-121. Supplementary Table 10 has a list of the causal SNPs. One thing that surprises me is how few of them are coding SNPs.

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  8. thank you for simplifying this!! btw, Razib Khan & Spencer Wells have a great interview with James Lee about this study on their podcast: http://insitome(dot)libsyn(dot)com/website

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter Display All Comments
  9. hyperbola says:
    @res

    It barely needs saying, but the tissues these SNPs most enrich are in the brain.
     
    And pituitary! (again) From page 121 of the Supplementary Material:

    When the tissues are ranked by heritability enrichment, the top result is frontal cortex (10.4-fold enrichment, P = 2.84×10–5), followed by pituitary (7.65-fold enrichment, P = 0.08) and broadly expressed (4.42-fold enrichment, P = 2.57×10–4). When the tissues are ranked by proportion of heritability, however, broadly expressed (proportion of h2 = 0.026, one-sided P = 6.25×10–7) overtakes frontal cortex (proportion of h2 = 0.019, one-sided P = 1.36×10–6). If we assume (unrealistically) no sampling covariance between these two estimates, then the difference between broadly expressed122 and frontal cortex in proportion of heritability is not significant (P = 0.30). The remaining results combined do not account for even half as much heritability as frontal cortex.
     
    Supplementary Figure 24 on page 198 shows this visually.

    Supplementary Figure 27 gives an interesting look at predictive power by chromosome. Chromosome 2 is both large and disproportionately predictive.

    P.S. Here is the caption for your lead figure:

    Supplementary Figure 8. Roles of Selected Newly Prioritized Genes in Neuronal Communication.
    The 59 genes listed in the figure were selected as follows. We began with the 30 gene-set clusters in Supplementary Figure 22 and dropped those that include gene sets that were implicated in a previous study of EduYears (Supplementary Table 4.5.1 of Okbay et al. 1). Of the 8 clusters that remained, we retained the 4 related to neuronal communication (DAG and IP3 signaling, associative learning, post NMDA receptor activation events, regulation of neurotransmitter levels).
    We identified the 460 DEPICTprioritized genes belonging to the exemplary gene sets representing these clusters (membership Z score > 2). Of these, the figure shows the 59 genes that appear in a figure or table of Fain352; these are genes whose functions are considered important for neuronal physiology
     

    This article has already been extensively discussed in two previous threads here at UNZ review. Now we have still more gobble-de-gook about a “paper” that needs 200 pages to hide the real conclusions.

    These “million person” studies are so under-sampled as to be nothing but random noise, which is why each one of them finds completely different sets of genes. Take 40 genes with 2 variants each (this is only a very small fraction of the genes that have been proposed to be correlated with things like “IQ”). The possible number of genetic variants is then 2 to the 40th power – that is more than ONE TRILLION combinations of those 40 genes with only two variants each. A sample size of one million persons corresponds to testing less than 1 part per million of those combinations.

    This leads to two inescapable conclusions.

    1. There are NOT enough people in the world to EVER do meaningful statistics on complex traits that involve >ca. 30 genes. ALL of these studies are measuring nothing more than noise artefacts.

    2. NO genetic test will ever be able to predict the “IQ” (or related things like “education achievement”) for an individual person.

    The question we should be asking ourselves is why we are being sold so much obvious snake oil from “psychologists”.

    Read More
    • Replies: @res
    Do you think if you repeat this enough times in enough places it will magically become true?

    Regarding 1. Do you understand how additive behavior works?

    Regarding 2. Do you believe the same is true of height?
    , @utu
    You have been bringing up this issue before. Have you thought of it since?

    Indeed there is a problem that the mathematical system in insanely undetermined. There are 10 millions SNPs that can be considered as discreet variables assuming value of 0,1 or 2 and n=1.1 million sample of dependent variable. If you constructed a linear function (polygenic score) out of all SNPs then obviously there is infinite number of solutions. Therefore the question must be asked differently: find the minimal number of SNPs that can be included in the polygenic score that explains the dependent variable. Minimum in what sense if for every subset of SNPs you can add extra SNP and probably improve prediction so when do you know when to stop? When do you know that you are not overfitting?

    You have to divide you sample into two samples: exploratory and validation subsets. On the exploratory set you determine coefficients of your function for a given subset of SNPs and on the validation set you measure the R^2. The R^2 on the exploratory set will be increasing when you will be adding SNPs to your polygenic score but at some point on the validation subset R^2 will stop increasing. This will mean that you have reached the point of overfitting on the exploratory set.

    Does this approach guarantee that you find the only one solution? No it is possible there are other subsets of SNPs that may produce similar or even better fits that are valid. The whole trick and mathematical difficulty is to find these subsets from, for all practical reason, of among the infinity of combinations. Steven Hsu hyped everybody with his Lasso (L1-fit) method but actually he "cheated" because he reduced the original set of all SNPs to these that had measurable correlation with the phenotype (dependent variable). So instead of millions of SNPs he ended up with 50,000 from which he found solution with 10,000 SNPs that explained 9% of educational attainment. Is it possible that there are other subsets outside his 50k subset that could also do the job? Yes, but we will never know. SNPs that do not have measurable correlation may correlate in groups so it means they should not be excluded from a possible solution as Hsu excluded them.

    Does the procedure using exploratory and validation samples establishes causality? I am pondering on this issue and do not have a good answer. The answer will be statistical, i.e, there will be a nonzero probability that a solution is not causal but spurious.The probability will depend on sizes of exploratory and validation samples.

    Does linear (additive) polygenic score is justified? I do not know. The problem is complex. When using linear polygenic score it becomes much simpler mathematically. But there are simplifications which I am not sure that are justified. Presumably they might be justified by GWAS. I do not know. For example a linear polygenic score implies that the additive effect from 0 allele to 1 allele change is the same as from 1 allele two 2 alleles. Then it implies that different SNPs are ignorant of each other while it might be possible that the effect of having SNP1 and SNP2 is not a sum of effects of the two SNPs together. To discover it one would hav to either do non-linear polygenic score or even better to an equivalent of logistic regression where SNPs cannot be mathematized to numerical values.

    Another issue related to sample size. Is possible that say n=1.1 million sample GWAS will indicate that SNP_x definitively is responsible for intelligence but when the sample is increased to 2.2 million it will be found out that SNP_x is not responsible for intelligence? Absolutely! That's why GWAS results re always sample depended and not universal.

    Is it possible that even if we have 7 billion sample of all humans on Earth the set of SNPs found by GWAS (1) will not be not complete or (2) will contain spurious not causal SNPs? Yes on both counts.
    , @utu
    Let me address one more time the issue of the undetermined system that you bring up. The issue is about how to determine whether a correlation produced by a given model is causal or spurious. The method requires that the model is developed separately on the exploratory sample and tested on the validation sample that must be independent of the exploratory sample. The validation sample may not be a subset of the exploratory sample!

    Let's imagine there is an unscrupulous scientist (Davide Piffer comes to my mind) who gets hold of 7 billion sample of IQ's and SNPs. He can use the whole sample as the exploratory sample and develops a model (polygenic score) that predicts IQ with R^2. Then he can extracts a subsample N<7 billion and call it a validation sample and pretends it is independent and he will get close to R^2 on this validation sample for his model. There will be no way of undoing his result. Everybody using his model will get the same results. There will be no way of determining whether the correlation R^2 is spurious or causal.

    This is a danger of dealing with severely undetermined mathematical systems. Can this be avoided? No. Sooner or later it will happen as different samples will be merged into one big mega sample and the "chain of custody" will be lost.
    , @puropedo
    Very good comment Hyperbola , psychologists , biologists , sociologists , pedagogues , physicians ,etc... are selling snake oil . They are selling a very expensive ,and materialistic " science " , which of course is a very good bussiness for them , but not so much for the general population who pays their academic delusions .
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  10. res says:
    @hyperbola
    This article has already been extensively discussed in two previous threads here at UNZ review. Now we have still more gobble-de-gook about a "paper" that needs 200 pages to hide the real conclusions.

    These “million person” studies are so under-sampled as to be nothing but random noise, which is why each one of them finds completely different sets of genes. Take 40 genes with 2 variants each (this is only a very small fraction of the genes that have been proposed to be correlated with things like “IQ”). The possible number of genetic variants is then 2 to the 40th power – that is more than ONE TRILLION combinations of those 40 genes with only two variants each. A sample size of one million persons corresponds to testing less than 1 part per million of those combinations.

    This leads to two inescapable conclusions.

    1. There are NOT enough people in the world to EVER do meaningful statistics on complex traits that involve >ca. 30 genes. ALL of these studies are measuring nothing more than noise artefacts.

    2. NO genetic test will ever be able to predict the “IQ” (or related things like “education achievement”) for an individual person.

    The question we should be asking ourselves is why we are being sold so much obvious snake oil from “psychologists”.

    Do you think if you repeat this enough times in enough places it will magically become true?

    Regarding 1. Do you understand how additive behavior works?

    Regarding 2. Do you believe the same is true of height?

    Read More
    • Replies: @hyperbola
    There are any number of medical conditions (rare diseases) where sample sizes much less than one million are adequate to define the genes responsible for the disease. Often one finds that one or a handful of genes that are both necessary and sufficient to cause the disease.

    What the million-sample "IQ" studies show is that there is NO small number of genes that are necessary and sufficient for producing such a complex (and very poorly defined) trait. For "IQ", THE RESULTS ARE IN (unless you wish to propose that an extremely improbable set of people have ended up in your million person sample).

    Waving magic solutions like "polygenic" (why should "additive" be linear - complex interdependence is physiologically much more reasonable) is simply an attempt to cover up for failure. If there were as few as 10 genes, the studies should already have given an overwhelmingly clear answer. Time to acknowledge that for "IQ" the situation is so complex that even with 30 genes essentially all of humanity would have to be measured to find those 30 genes.
    , @Okechukwu

    Regarding 2. Do you believe the same is true of height?
     
    Leave it to resident IQ-obsessed troll res to liken something as diatarily regulated as height to the most complex trait we know -- a trait that science still doesn't understand and maybe never will.

    The human brain is literally as opaque and as indecipherable as the universe. On the other hand, eat your vegetables and drink your milk and you'll grow taller.

    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  11. Illiterate thinks a nonsense abstraction like “intelligence” is the same as “height.”

    And such people are taken seriously.

    More lies, damned lies, and you know whats.

    Read More
    • LOL: res
    ReplyAgree/Disagree/Etc. More... This Commenter Display All Comments
  12. “It would be good to have far more genetic and intelligence data on Africans . . .”

    It would be far better to locate the genetic markers that uniquely identify intelligence in humans if it exists.

    But I appreciate the straining of gnats.

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter Display All Comments
  13. Anonymous[321] • Disclaimer says:
    @James Thompson
    Yes, the pituitary is trying to tell us something. Thanks for the more detailed list, which I skipped over.

    Now this is interesting, and I wonder if the pineal gland is involved too. As far as I know, these two do not have neurons but act by secreting hormones, which means their influence would appear to be indirect.

    This perked my interest because these two glands are traditionally held to be the physical correlations of the 6th and 7th chakras which are involved in attaining higher levels of consciousness. Higher consciousness grants clearer perception, which can be measured as higher intelligence.

    So then it seems we’re not just talking about the correlation of physical structure and intelligence. Instead, we’re talking about a three-way correlation: consciousness, intelligence, and physical structure.

    Is it allowed to bring consciousness into this, James Thompson? :-)

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  14. It is all well and good to admit that blacks have an IQ on average perhaps 10 points lower than white Christians (100-105) but then we’d have to admit that Jews have an IQ of 7 points greater than whites and a good number of them are NASA brainy. Einstein, for example.

    Now don’t get me wrong, Jews are a tiny minority and some other factor (Nepotism, geography, tradition) besides IQ must matter in a population of so many white Americans that have sky-high IQ’s.

    But the fact of the matter is that the white prole does STUPID things and much of this is related to environment. Are you a genius whose parents were hillbilly transplants to Detroit like Eminem who found yourself in a Detroit public school? If you’re smart, you’ll be a whigger. But only one in millions of whiggers succeeds like Eminem. If you’re a girl with a JAP or Asian IQ in an integrated school, as one other poster mentioned, you’ll succumb to sexual bullying and might end up with a fatherless child.

    More often, in my experience, the smart kids end up being stoners. They get into pot in junior high, maybe as early as the 7th grade and by age 16 they are into crystal meth (Rich kids can afford cocaine and rehab) and at this point they commit the same petty crimes as blacks to get it-the burglary, the property theft, the cons, the dealing. Most whites who end up in prison are not really as evil as Jeffrey Dahmer or Tony Soprano-they just got pulled over with burglary tools and possession of meth with intent.

    The lucky lower-income whites are the hicks who live in small towns. Nobody gets jumped at the 4-H club meetings by blacks who usually mature quicker than whites anyhow.

    But being an intelligent hick in a small town has its limitations. Perhaps your parents can afford for you to attend college but a state school, not Ivy League. Plus, rural whites tend to marry young (Catholics and Lutherans less) and so these brainy whites marry at 21, have a kid about that age and economic pressure leads them away from scholastic achievement or big money.

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter Display All Comments
  15. FB says:

    What a massive load of crapola…

    Here is the proof…

    ‘…Research has now moved to an international dimension, with disparate groups being managed and cajoled into cooperative ventures, a major undertaking that requires academia to develop new skills of diplomacy, coordination of disparate research groups, careful assembly of very different data formats and research protocols, and a sensitive understanding of individual egos and conflicting cultural and political sensitivities…

    So that’s how its done in the social ‘sciences’…

    Funny because in real science international cooperation has been a ‘thing’ now for about OH SINCE ABOUT NEWTON…

    LOL…

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter Display All Comments
  16. utu says:

    How many SNPs are in the polygenic score? How this polygenic score is constructed? In the paper we read.

    All scores are based on the results from a meta-analysis that excluded the prediction cohorts. Our first four scores were con-structed from sets of LD-pruned SNPs associated with EduYears at various P-value thresholds: 5 ×10−8, 5 ×10−5, 5 ×10−3 and 1 (that is, all SNPs). In both cohorts, the predictive power is greater for scores constructed with less stringent thresholds (Supplementary Fig. 10). The sample-size-weighted mean incremental R2 increases from 3.2% at P < 5 × 10−8 to 9.4% at P ≤ 1. Our fifth score was generated from HapMap3 SNPs using the software LDpred26. Rather than removing SNPs that are in LD with each other, LDpred is a Bayesian method that weights each SNP by (an approximation to) the posterior mean of its conditional effect, given other SNPs. This score was the most predic-tive in both cohorts, with an incremental R2 of 12.7% in AddHealth and 10.6% in HRS (and a sample-size weighted mean of 11.4%)

    Depending on how many SNPs were removed because of criteria on LD their prediction (incremental variance explained) is 3.2%, 9.4% and 12.7%. The best case is when SNPs that are in LD are not removed but weighted in the score. So where do the difference between 3.2% and 12.7% come from? Let’s first go to their summary where they talk what was really accomplished:

    For research at the intersection of genetics and neuroscience, the set of 1,271 lead SNPs that we identify is a treasure trove for future analyses. For research in social science and epidemiology, the polygenic scores that we construct—which explain 11–13% and 7–10% of the variance in educational attainment and cognitive per-formance, respectively—will prove useful across at least three types of applications.

    This summary suggests that with just 1,271 SNPs they can explain 11-13% of education attainment variance. But is it true? They call these SNPs a treasure trove for future analyses. Let’s go to FAQs notes for more explanation:

    Taken together, these 1,271 SNPs accounted for just 3.9% of the variation across individuals in years of education completed.

    As discussed in FAQ 1.5, we can create an index using the GWAS results from around ~1 million genetic variants. Such an index is called a “polygenic score.”

    The polygenic score we constructed “predicts” (see FAQ 1.4) around 11% of the variation in education across individuals (when tested in independent data that was not included in the GWAS). This ~1 million SNP polygenic score predicts much more of the variation than does the genetic predictor described in FAQ 2.2, which was based on only 1,271 SNPs. Including all ~1 million SNPs tends to add predictive power because the threshold for significance/inclusion that is used to identify the 1,271 SNPs is very conservative (i.e., many of the other ~1 million SNPs are also associated with educational attainment but are not identified by our study, and on net, it turns out empirically that more signal than noise is added by including them).

    The truth is which does not jump at you form the main body of the paper is that to obtain 11-13% prediction each “lead SNP” from the “treasure trove” of 1,271 has to be combined with additional 787 SNPs on average. ONE MILLION SPNs in the polygenic score is a lot. The polygenic score uses 10% of all SNPs there are in human genome.

    What about the sample size? The 1.1 million sample is exploratory sample on which GWAS is conducted and weights for each SNP in the polygenic score are determined if I understand it correctly. The prediction (validation) samples are much smaller: Adolescent to Adult Health (Add Health, n = 4,775) and Health and Retirement Study (HRS, n = 8,609).

    Let suppose we remove 1,271 “lead SNPs” from the polygenic score. Does it mean that remaining 1 million-1,271 SPNs predict 11%-3.9%=7.1%?

    Does this paper deserve the hype it gets?

    Are the authors guilty of some obfuscation?

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter Display All Comments
  17. utu says:
    @hyperbola
    This article has already been extensively discussed in two previous threads here at UNZ review. Now we have still more gobble-de-gook about a "paper" that needs 200 pages to hide the real conclusions.

    These “million person” studies are so under-sampled as to be nothing but random noise, which is why each one of them finds completely different sets of genes. Take 40 genes with 2 variants each (this is only a very small fraction of the genes that have been proposed to be correlated with things like “IQ”). The possible number of genetic variants is then 2 to the 40th power – that is more than ONE TRILLION combinations of those 40 genes with only two variants each. A sample size of one million persons corresponds to testing less than 1 part per million of those combinations.

    This leads to two inescapable conclusions.

    1. There are NOT enough people in the world to EVER do meaningful statistics on complex traits that involve >ca. 30 genes. ALL of these studies are measuring nothing more than noise artefacts.

    2. NO genetic test will ever be able to predict the “IQ” (or related things like “education achievement”) for an individual person.

    The question we should be asking ourselves is why we are being sold so much obvious snake oil from “psychologists”.

    You have been bringing up this issue before. Have you thought of it since?

    Indeed there is a problem that the mathematical system in insanely undetermined. There are 10 millions SNPs that can be considered as discreet variables assuming value of 0,1 or 2 and n=1.1 million sample of dependent variable. If you constructed a linear function (polygenic score) out of all SNPs then obviously there is infinite number of solutions. Therefore the question must be asked differently: find the minimal number of SNPs that can be included in the polygenic score that explains the dependent variable. Minimum in what sense if for every subset of SNPs you can add extra SNP and probably improve prediction so when do you know when to stop? When do you know that you are not overfitting?

    You have to divide you sample into two samples: exploratory and validation subsets. On the exploratory set you determine coefficients of your function for a given subset of SNPs and on the validation set you measure the R^2. The R^2 on the exploratory set will be increasing when you will be adding SNPs to your polygenic score but at some point on the validation subset R^2 will stop increasing. This will mean that you have reached the point of overfitting on the exploratory set.

    Does this approach guarantee that you find the only one solution? No it is possible there are other subsets of SNPs that may produce similar or even better fits that are valid. The whole trick and mathematical difficulty is to find these subsets from, for all practical reason, of among the infinity of combinations. Steven Hsu hyped everybody with his Lasso (L1-fit) method but actually he “cheated” because he reduced the original set of all SNPs to these that had measurable correlation with the phenotype (dependent variable). So instead of millions of SNPs he ended up with 50,000 from which he found solution with 10,000 SNPs that explained 9% of educational attainment. Is it possible that there are other subsets outside his 50k subset that could also do the job? Yes, but we will never know. SNPs that do not have measurable correlation may correlate in groups so it means they should not be excluded from a possible solution as Hsu excluded them.

    Does the procedure using exploratory and validation samples establishes causality? I am pondering on this issue and do not have a good answer. The answer will be statistical, i.e, there will be a nonzero probability that a solution is not causal but spurious.The probability will depend on sizes of exploratory and validation samples.

    Does linear (additive) polygenic score is justified? I do not know. The problem is complex. When using linear polygenic score it becomes much simpler mathematically. But there are simplifications which I am not sure that are justified. Presumably they might be justified by GWAS. I do not know. For example a linear polygenic score implies that the additive effect from 0 allele to 1 allele change is the same as from 1 allele two 2 alleles. Then it implies that different SNPs are ignorant of each other while it might be possible that the effect of having SNP1 and SNP2 is not a sum of effects of the two SNPs together. To discover it one would hav to either do non-linear polygenic score or even better to an equivalent of logistic regression where SNPs cannot be mathematized to numerical values.

    Another issue related to sample size. Is possible that say n=1.1 million sample GWAS will indicate that SNP_x definitively is responsible for intelligence but when the sample is increased to 2.2 million it will be found out that SNP_x is not responsible for intelligence? Absolutely! That’s why GWAS results re always sample depended and not universal.

    Is it possible that even if we have 7 billion sample of all humans on Earth the set of SNPs found by GWAS (1) will not be not complete or (2) will contain spurious not causal SNPs? Yes on both counts.

    Read More
    • Replies: @hyperbola
    I addressed this a bit more in the answer to res - I maintain that we have already proven that there is NO small set of genes that can explain the results. To me the important question is this: if I know the exact genetic makeup of an individual, can I predict his "IQ" (or related trait like educational achievement - both of which are very poorly defined). If the complexity of the trait is so large that only a very vague average statistical classification is possible, then the "psychologists" are cooking up very ugly and dangerous "social" poisons. Given the historical development of "modern psychology", this would not surprise me.

    I understand the sample size problem and value your comments. Perhaps one might address this topic from a different viewpoint: how many individuals can I find with exactly the same set of SNPs and what is the variance of their "IQ"?
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  18. ” This paper covers considerable ground, and does so in a careful way. I think it will have a massive impact. ”
    Why a massive impact, and what impact ?

    Read More
    • Replies: @Impact
    I think that they mean that they will publish the paper in a " prestigious " magazine that has a high index of impact , say 5 ,10 15 points or more per publication .... .

    They have a ranking of scientific magazines , english only of course , and each magazine is given an index of impact , say publishing in " the crazy neuron " 5 points of impact , publishing in " the crazy neurotransmiter " 10 points if impact , publishing in " the crazy gene " 50 points of impact etc....

    If the group of research has a lot of impact , " a massive impact " , their academic careers will advance a lot , and they will be able to file for more grants to keep their research going . Nowadays a University Professor without impact in nothing ,

    Often you have to pay to publish in high impact publications . How come that we have this situation in the era of free internet ? . Well , in the area of biosciences all or nearly all the " prestigious " few high impact magazines are controlled by 6 publishing companies , by the way one of then dutch , they publish only in english , thus marginalizing other languages , and they consider only the prevalent scientific paradigm which is represented by the paper object ot the article , disidence or other scientific approaches or conceptualizations are margizalized .
    , @NIH
    https://en.wikipedia.org/wiki/National_Institutes_of_Health

    the biggest biomedical research institution of the world , the US Government : National Institute of Health ( NIH ) , 2018 budget 37.000 million dollars .
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  19. utu says: • Website
    @hyperbola
    This article has already been extensively discussed in two previous threads here at UNZ review. Now we have still more gobble-de-gook about a "paper" that needs 200 pages to hide the real conclusions.

    These “million person” studies are so under-sampled as to be nothing but random noise, which is why each one of them finds completely different sets of genes. Take 40 genes with 2 variants each (this is only a very small fraction of the genes that have been proposed to be correlated with things like “IQ”). The possible number of genetic variants is then 2 to the 40th power – that is more than ONE TRILLION combinations of those 40 genes with only two variants each. A sample size of one million persons corresponds to testing less than 1 part per million of those combinations.

    This leads to two inescapable conclusions.

    1. There are NOT enough people in the world to EVER do meaningful statistics on complex traits that involve >ca. 30 genes. ALL of these studies are measuring nothing more than noise artefacts.

    2. NO genetic test will ever be able to predict the “IQ” (or related things like “education achievement”) for an individual person.

    The question we should be asking ourselves is why we are being sold so much obvious snake oil from “psychologists”.

    Let me address one more time the issue of the undetermined system that you bring up. The issue is about how to determine whether a correlation produced by a given model is causal or spurious. The method requires that the model is developed separately on the exploratory sample and tested on the validation sample that must be independent of the exploratory sample. The validation sample may not be a subset of the exploratory sample!

    Let’s imagine there is an unscrupulous scientist (Davide Piffer comes to my mind) who gets hold of 7 billion sample of IQ’s and SNPs. He can use the whole sample as the exploratory sample and develops a model (polygenic score) that predicts IQ with R^2. Then he can extracts a subsample N<7 billion and call it a validation sample and pretends it is independent and he will get close to R^2 on this validation sample for his model. There will be no way of undoing his result. Everybody using his model will get the same results. There will be no way of determining whether the correlation R^2 is spurious or causal.

    This is a danger of dealing with severely undetermined mathematical systems. Can this be avoided? No. Sooner or later it will happen as different samples will be merged into one big mega sample and the "chain of custody" will be lost.

    Read More
    • Replies: @hyperbola
    There may be means to test whether any particular SNP is causal. For example, for rare diseases (where only one or a small handful of genes are involved) there are now attempts to verify and "fix" the medical problem by altering the gene. This type of approach probably becomes too onerous and uninteresting when hundreds of genes are involved.

    By the way, from medicine we know that the vast majority of "disease" conditions involve coding regions of the genome. This is why I prefer to speak about genes rather than SNPs. It also potentially reduces the size of the problem to about 19,000 genes rather than millions of SNPs.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  20. puropedo says:
    @hyperbola
    This article has already been extensively discussed in two previous threads here at UNZ review. Now we have still more gobble-de-gook about a "paper" that needs 200 pages to hide the real conclusions.

    These “million person” studies are so under-sampled as to be nothing but random noise, which is why each one of them finds completely different sets of genes. Take 40 genes with 2 variants each (this is only a very small fraction of the genes that have been proposed to be correlated with things like “IQ”). The possible number of genetic variants is then 2 to the 40th power – that is more than ONE TRILLION combinations of those 40 genes with only two variants each. A sample size of one million persons corresponds to testing less than 1 part per million of those combinations.

    This leads to two inescapable conclusions.

    1. There are NOT enough people in the world to EVER do meaningful statistics on complex traits that involve >ca. 30 genes. ALL of these studies are measuring nothing more than noise artefacts.

    2. NO genetic test will ever be able to predict the “IQ” (or related things like “education achievement”) for an individual person.

    The question we should be asking ourselves is why we are being sold so much obvious snake oil from “psychologists”.

    Very good comment Hyperbola , psychologists , biologists , sociologists , pedagogues , physicians ,etc… are selling snake oil . They are selling a very expensive ,and materialistic ” science ” , which of course is a very good bussiness for them , but not so much for the general population who pays their academic delusions .

    Read More
    • Replies: @hyperbola
    There is good science and fraudulent science. Unfortunately "our" society now provides vastly more motivation to fraudulent scientists than to honest scientists. A related problem is the use of "pseudo-science" to sell "products".
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  21. APilgrim says:

    About ‘our rootstock’ …

    Neanderthal genes, in Northern Hemisphere populations are directly tied to IQ. Caucasians & Asians have average IQs of over 100. Black Africans & Australian Aboriginals have average IQs below 70. Native Middle Eastern populations and American ‘coloreds’ have average IQs of approximately 85.

    Neanderthal brains were on average larger than those of fully modern humans, and research points toward symbolic thought.

    Read More
    • Replies: @Okechukwu

    Neanderthal genes, in Northern Hemisphere populations are directly tied to IQ.
     
    Neanderthal was a glorified gorilla. Neanderthal admixture is a minus, not a plus.

    Caucasians & Asians have average IQs of over 100. Black Africans & Australian Aboriginals have average IQs below 70.
     
    Australian Aborigines carry Neanderthal DNA.

    Caucasians include MENA and Indian sub-continent people. They do not have an average IQ over 100. In fact black Americans have higher IQ's, as do a large portion black Africans. That's assuming tha any of this IQ gobbledygook is even real. I think it's safe to say that IQ is one of the biggest hoaxes ever foisted upon humanity.

    By the way, if Africans have an IQ below 70, which would qualify them for the Special Olympics, then you'd better come up with a plausible explanation as to why I have yet to encounter one as dumb as you are.


    Neanderthal brains were on average larger than those of fully modern humans, and research points toward symbolic thought.
     
    Research doesn't have to point toward symbolic thought among early humans. It's a given.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  22. APilgrim says:

    A SNIP Backgrounder:

    Single nucleotide polymorphisms, frequently called SNPs (pronounced “snips”), are the most common type of genetic variation among people. Each SNP represents a difference in a single DNA building block, called a nucleotide. For example, a SNP may replace the nucleotide cytosine (C) with the nucleotide thymine (T) in a certain stretch of DNA.

    SNPs occur normally throughout a person’s DNA. They occur once in every 300 nucleotides on average, which means there are roughly 10 million SNPs in the human genome. Most commonly, these variations are found in the DNA between genes. They can act as biological markers, helping scientists locate genes that are associated with disease. When SNPs occur within a gene or in a regulatory region near a gene, they may play a more direct role in disease by affecting the gene’s function.

    Most SNPs have no effect on health or development. Some of these genetic differences, however, have proven to be very important in the study of human health. Researchers have found SNPs that may help predict an individual’s response to certain drugs, susceptibility to environmental factors such as toxins, and risk of developing particular diseases. SNPs can also be used to track the inheritance of disease genes within families. Future studies will work to identify SNPs associated with complex diseases such as heart disease, diabetes, and cancer.

    Read More
    • Replies: @Lost american
    APilgrim- thanks for giving a good explanation that most people can follow. I worked in science, excelled in standardized tests in my field.
    I always felt that the best teachers could simplify things so that the person on the street could better understand.
    I have a disdain for those who like to use language and symbols way beyond the expertise of over 90% reading an article such as this one. Most people that use Unz Review are concerned, perceptive individuals. Most are not geniuses nor are they in the superior IQ category, but they have a strong desire to learn why things tick whether in science or in the way societies operate.
    Some people seem to want others to be impressed so they write in a way that few can understand. Perhaps it makes their swollen egos swell a lot more. I bet that most of these people with big egos would be lost if they had to connect pipes that one learns in plumbing 102.
    I am getting off track. Thanks for explaining things so that all types can understand the topic such as SNPs.
    , @Tim too
    The text is quote from:
    https://ghr.nlm.nih.gov/primer/genomicresearch/snp

    should be attributed.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  23. anon[317] • Disclaimer says:

    I think the paper should have been published for its data set. and the documentation of the statistical methodologies used, and the current technologies in software and data analysis described, but I am not so sure I can accept its conclusion or its inferential suggestions. And to make its publisher a lot of money.

    Not until the entire set of biological processes that support neural controlled behavior in living creatures in always changing environments are complied into a single interactive dynamic, input controllable model capable to simulate results that express human mental behaviors as functions of the inputs will I believe reliable predictive power has been achieved.

    thanks for the paper, it keep me busy last evening.

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter Display All Comments
  24. Impact says:
    @jilles dykstra
    " This paper covers considerable ground, and does so in a careful way. I think it will have a massive impact. "
    Why a massive impact, and what impact ?

    I think that they mean that they will publish the paper in a ” prestigious ” magazine that has a high index of impact , say 5 ,10 15 points or more per publication …. .

    They have a ranking of scientific magazines , english only of course , and each magazine is given an index of impact , say publishing in ” the crazy neuron ” 5 points of impact , publishing in ” the crazy neurotransmiter ” 10 points if impact , publishing in ” the crazy gene ” 50 points of impact etc….

    If the group of research has a lot of impact , ” a massive impact ” , their academic careers will advance a lot , and they will be able to file for more grants to keep their research going . Nowadays a University Professor without impact in nothing ,

    Often you have to pay to publish in high impact publications . How come that we have this situation in the era of free internet ? . Well , in the area of biosciences all or nearly all the ” prestigious ” few high impact magazines are controlled by 6 publishing companies , by the way one of then dutch , they publish only in english , thus marginalizing other languages , and they consider only the prevalent scientific paradigm which is represented by the paper object ot the article , disidence or other scientific approaches or conceptualizations are margizalized .

    Read More
    • Replies: @jilles dykstra
    I see.
    The same as in climate sciences ?
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  25. NIH says:
    @jilles dykstra
    " This paper covers considerable ground, and does so in a careful way. I think it will have a massive impact. "
    Why a massive impact, and what impact ?

    https://en.wikipedia.org/wiki/National_Institutes_of_Health

    the biggest biomedical research institution of the world , the US Government : National Institute of Health ( NIH ) , 2018 budget 37.000 million dollars .

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  26. pyrrhus says: • Website

    Fascinating post, thank you Dr. Thompson! One comment about methodology….
    The correlation between pure intelligence and educational attainment must be much better in Europe than it is in the US…These days in America, virtually everyone, including students with below average IQs, can and does take out student loans and enroll in some kind of open admission, often for-profit, institution of higher “learning.” Unfortunately, this generally results in significant debt for no economic benefit, and a level of educational “attainment” unconnected to IQ…We also have lots of scholarship athletes, many of them African-American, who graduate with a 4 year degree but are functionally illiterate.
    On the other end, I have known or taught very bright students who dropped out of college, and even high school, generally to work full time in the IT field or gaming.

    Read More
    • Replies: @JackOH
    " . . . [A] level of educational “attainment” [in the States] unconnected to IQ . . .".

    Exactly so, Pyrrhus. I'm a close observer of my local less selective state university, and it's my well-informed guess that 80% of the students derive no benefit from attendance, and, what's worse, there may be actual damage done to these students. They'll have degrees conferred upon them that will mislead many into believing they're much more able than they actually are.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  27. res says:

    Some of the commenters here might find Eric Turkheimer’s take on EA3 more to their liking: http://www.geneticshumanagency.org/gha/ea3-better-formatting/

    Except, no comments allowed ; )

    I was especially impressed by Turkheimer’s exceptional modesty:

    I should add that it is first authored by GHA participant James Lee, so I take most of the credit for it.

    I guess we are just about to the phase where everyone says “that was obvious” despite having denied exactly that (e.g. hate for The Bell Curve) for decades:

    On the other hand, it must be remembered that the basic theoretical insight that genes have a role to play in human differences like intelligence and education has been well-established for decades.

    One good thing is he linked to Robert Plomin’s forthcoming book: https://mitpress.mit.edu/books/blueprint

    Read More
    • Replies: @James Thompson
    Thanks for directing me to Turkheimer's remarks, which end thus:

    it does not show that differences in outcomes among racial and ethnic groups are genetically determined. In fact none of these things are any more likely today than they were yesterday.

    I think that this is wrong, in that every paper which shows a detailed link between the genetic code of one genetic group and an important outcome, like years of completed education, strengthens the likelihood that another genetic group might have a different code which leads to different outcomes. However, if it was impossible to show that there was any link in any genetic group with educational outcomes, then the chance of it accounting for racial differences in educational attainment would be much weaker. Currently, searching for genetic differences as an explanation for genetic group outcomes has become more tenable, not less.
    , @Okechukwu

    I guess we are just about to the phase where everyone says “that was obvious” despite having denied exactly that (e.g. hate for The Bell Curve) for decades:
     
    How misleading and deceptive. Nothing in the paper you linked is an endorsement of The Bell Curve. To this very moment, Turkheimer stands behind his criticism of that book. Don't take my word for it, Email him.

    Charles Murray is once again peddling junk science about race and IQ

    https://www.vox.com/the-big-idea/2017/5/18/15655638/charles-murray-race-iq-sam-harris-science-free-speech


    On the other hand, it must be remembered that the basic theoretical insight that genes have a role to play in human differences like intelligence and education has been well-established for decades.
     
    How exactly is this controversial? And how does it go to support The Bell Curve? Notice Turkheimer's reference is to human (i.e., individual differences) and not group differences. His twins citation could be twins of any race or color. If someone grows up in my extended family they're probably going to be high achievers educationally because that's just what we do.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  28. Sorry .. but I consider the Article RUBBISH … It isnt able to tell normal people like me what it all is about ..Aleegedly it throws several diverse studies in a Pot and extracts some ” conclusions from this Soup … it hides itself behing cryptic abbreviations and hard to look through pseudo statistic mumbo jumbo … Sorry Folks .. The Emperor has no clothes !

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter Display All Comments
  29. @res
    Some of the commenters here might find Eric Turkheimer's take on EA3 more to their liking: http://www.geneticshumanagency.org/gha/ea3-better-formatting/

    Except, no comments allowed ; )

    I was especially impressed by Turkheimer's exceptional modesty:

    I should add that it is first authored by GHA participant James Lee, so I take most of the credit for it.
     
    I guess we are just about to the phase where everyone says "that was obvious" despite having denied exactly that (e.g. hate for The Bell Curve) for decades:

    On the other hand, it must be remembered that the basic theoretical insight that genes have a role to play in human differences like intelligence and education has been well-established for decades.
     
    One good thing is he linked to Robert Plomin's forthcoming book: https://mitpress.mit.edu/books/blueprint

    Thanks for directing me to Turkheimer’s remarks, which end thus:

    it does not show that differences in outcomes among racial and ethnic groups are genetically determined. In fact none of these things are any more likely today than they were yesterday.

    I think that this is wrong, in that every paper which shows a detailed link between the genetic code of one genetic group and an important outcome, like years of completed education, strengthens the likelihood that another genetic group might have a different code which leads to different outcomes. However, if it was impossible to show that there was any link in any genetic group with educational outcomes, then the chance of it accounting for racial differences in educational attainment would be much weaker. Currently, searching for genetic differences as an explanation for genetic group outcomes has become more tenable, not less.

    Read More
    • Replies: @hyperbola
    Unfortunately the pseudo-science in this paper does NOT support your conclusion.

    Currently, searching for genetic differences as an explanation for genetic group outcomes has become more tenable, not less.
     
    It seems to be time to exclude the "psychologists" from being examples of science.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  30. @Impact
    I think that they mean that they will publish the paper in a " prestigious " magazine that has a high index of impact , say 5 ,10 15 points or more per publication .... .

    They have a ranking of scientific magazines , english only of course , and each magazine is given an index of impact , say publishing in " the crazy neuron " 5 points of impact , publishing in " the crazy neurotransmiter " 10 points if impact , publishing in " the crazy gene " 50 points of impact etc....

    If the group of research has a lot of impact , " a massive impact " , their academic careers will advance a lot , and they will be able to file for more grants to keep their research going . Nowadays a University Professor without impact in nothing ,

    Often you have to pay to publish in high impact publications . How come that we have this situation in the era of free internet ? . Well , in the area of biosciences all or nearly all the " prestigious " few high impact magazines are controlled by 6 publishing companies , by the way one of then dutch , they publish only in english , thus marginalizing other languages , and they consider only the prevalent scientific paradigm which is represented by the paper object ot the article , disidence or other scientific approaches or conceptualizations are margizalized .

    I see.
    The same as in climate sciences ?

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  31. hyperbola says:
    @res
    Do you think if you repeat this enough times in enough places it will magically become true?

    Regarding 1. Do you understand how additive behavior works?

    Regarding 2. Do you believe the same is true of height?

    There are any number of medical conditions (rare diseases) where sample sizes much less than one million are adequate to define the genes responsible for the disease. Often one finds that one or a handful of genes that are both necessary and sufficient to cause the disease.

    What the million-sample “IQ” studies show is that there is NO small number of genes that are necessary and sufficient for producing such a complex (and very poorly defined) trait. For “IQ”, THE RESULTS ARE IN (unless you wish to propose that an extremely improbable set of people have ended up in your million person sample).

    Waving magic solutions like “polygenic” (why should “additive” be linear – complex interdependence is physiologically much more reasonable) is simply an attempt to cover up for failure. If there were as few as 10 genes, the studies should already have given an overwhelmingly clear answer. Time to acknowledge that for “IQ” the situation is so complex that even with 30 genes essentially all of humanity would have to be measured to find those 30 genes.

    Read More
    • Replies: @j2
    hyperbola writes:
    "(why should “additive” be linear – complex interdependence is physiologically much more reasonable)"

    This is a mathematical issue. Most of these complex interdependences are continuous, differentiable functions. A small effect is then f(p)+f '(p)dx, it is linear (i.e., effects are additive). This is true for small effects and it explains some part why such additive constructs like a polygenic score cannot explain much of the effect: it is additive (=linear) only for small changes. In order to find out the whole function (all partial derivatives) would require very much study.

    The mathematically more reasonable way to evaluate the effect of SNPs to IQ would be to show that an SNP has a small effect rather than no effect (as we cannot find out the whole effect without knowing the function) and then scan over all genes. The set of genes that have any effect account for the whole genetic effect, estimated to be 60-80%. Then you can assume that the effects are distributed in some way, like normally (a few have larger effects, most intermediate) and to calculate the average effect.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  32. hyperbola says:
    @utu
    You have been bringing up this issue before. Have you thought of it since?

    Indeed there is a problem that the mathematical system in insanely undetermined. There are 10 millions SNPs that can be considered as discreet variables assuming value of 0,1 or 2 and n=1.1 million sample of dependent variable. If you constructed a linear function (polygenic score) out of all SNPs then obviously there is infinite number of solutions. Therefore the question must be asked differently: find the minimal number of SNPs that can be included in the polygenic score that explains the dependent variable. Minimum in what sense if for every subset of SNPs you can add extra SNP and probably improve prediction so when do you know when to stop? When do you know that you are not overfitting?

    You have to divide you sample into two samples: exploratory and validation subsets. On the exploratory set you determine coefficients of your function for a given subset of SNPs and on the validation set you measure the R^2. The R^2 on the exploratory set will be increasing when you will be adding SNPs to your polygenic score but at some point on the validation subset R^2 will stop increasing. This will mean that you have reached the point of overfitting on the exploratory set.

    Does this approach guarantee that you find the only one solution? No it is possible there are other subsets of SNPs that may produce similar or even better fits that are valid. The whole trick and mathematical difficulty is to find these subsets from, for all practical reason, of among the infinity of combinations. Steven Hsu hyped everybody with his Lasso (L1-fit) method but actually he "cheated" because he reduced the original set of all SNPs to these that had measurable correlation with the phenotype (dependent variable). So instead of millions of SNPs he ended up with 50,000 from which he found solution with 10,000 SNPs that explained 9% of educational attainment. Is it possible that there are other subsets outside his 50k subset that could also do the job? Yes, but we will never know. SNPs that do not have measurable correlation may correlate in groups so it means they should not be excluded from a possible solution as Hsu excluded them.

    Does the procedure using exploratory and validation samples establishes causality? I am pondering on this issue and do not have a good answer. The answer will be statistical, i.e, there will be a nonzero probability that a solution is not causal but spurious.The probability will depend on sizes of exploratory and validation samples.

    Does linear (additive) polygenic score is justified? I do not know. The problem is complex. When using linear polygenic score it becomes much simpler mathematically. But there are simplifications which I am not sure that are justified. Presumably they might be justified by GWAS. I do not know. For example a linear polygenic score implies that the additive effect from 0 allele to 1 allele change is the same as from 1 allele two 2 alleles. Then it implies that different SNPs are ignorant of each other while it might be possible that the effect of having SNP1 and SNP2 is not a sum of effects of the two SNPs together. To discover it one would hav to either do non-linear polygenic score or even better to an equivalent of logistic regression where SNPs cannot be mathematized to numerical values.

    Another issue related to sample size. Is possible that say n=1.1 million sample GWAS will indicate that SNP_x definitively is responsible for intelligence but when the sample is increased to 2.2 million it will be found out that SNP_x is not responsible for intelligence? Absolutely! That's why GWAS results re always sample depended and not universal.

    Is it possible that even if we have 7 billion sample of all humans on Earth the set of SNPs found by GWAS (1) will not be not complete or (2) will contain spurious not causal SNPs? Yes on both counts.

    I addressed this a bit more in the answer to res – I maintain that we have already proven that there is NO small set of genes that can explain the results. To me the important question is this: if I know the exact genetic makeup of an individual, can I predict his “IQ” (or related trait like educational achievement – both of which are very poorly defined). If the complexity of the trait is so large that only a very vague average statistical classification is possible, then the “psychologists” are cooking up very ugly and dangerous “social” poisons. Given the historical development of “modern psychology”, this would not surprise me.

    I understand the sample size problem and value your comments. Perhaps one might address this topic from a different viewpoint: how many individuals can I find with exactly the same set of SNPs and what is the variance of their “IQ”?

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  33. hyperbola says:
    @utu
    Let me address one more time the issue of the undetermined system that you bring up. The issue is about how to determine whether a correlation produced by a given model is causal or spurious. The method requires that the model is developed separately on the exploratory sample and tested on the validation sample that must be independent of the exploratory sample. The validation sample may not be a subset of the exploratory sample!

    Let's imagine there is an unscrupulous scientist (Davide Piffer comes to my mind) who gets hold of 7 billion sample of IQ's and SNPs. He can use the whole sample as the exploratory sample and develops a model (polygenic score) that predicts IQ with R^2. Then he can extracts a subsample N<7 billion and call it a validation sample and pretends it is independent and he will get close to R^2 on this validation sample for his model. There will be no way of undoing his result. Everybody using his model will get the same results. There will be no way of determining whether the correlation R^2 is spurious or causal.

    This is a danger of dealing with severely undetermined mathematical systems. Can this be avoided? No. Sooner or later it will happen as different samples will be merged into one big mega sample and the "chain of custody" will be lost.

    There may be means to test whether any particular SNP is causal. For example, for rare diseases (where only one or a small handful of genes are involved) there are now attempts to verify and “fix” the medical problem by altering the gene. This type of approach probably becomes too onerous and uninteresting when hundreds of genes are involved.

    By the way, from medicine we know that the vast majority of “disease” conditions involve coding regions of the genome. This is why I prefer to speak about genes rather than SNPs. It also potentially reduces the size of the problem to about 19,000 genes rather than millions of SNPs.

    Read More
    • Replies: @utu

    This is why I prefer to speak about genes rather than SNPs.
     
    If this is true "that over 98% of the human genome does not encode protein sequences" (wiki) then out of 10 million SNPs only 200,000 belong to the coding regions. In this paper (See my comment #16) to obtain 11-13% variance explanation they effectively used 1 million of SNPs while with 1,271 "lead" SNPs they explained only 3.9%. This means that 800,000 SNPs they used in their polygenic score were form the noncoding regions. Not surprisingly commenter res wrote in his comment #6:

    Supplementary Table 10 has a list of the causal SNPs. One thing that surprises me is how few of them are coding SNPs.
     
    (BTW, where the Supplementary material is available?)

    It is possibly that the dichotomy of coding/non-coding DNA is false to some extent and there is much more going on than what was assumed so far in genetics. But if it was true that only SNP's from the coding regions should suffice to explain any trait, then if all 200,000 SNPs from coding regions were already included in the polygenic score and they still could not replicate the twin based heritability of some trait, then the case should be closed with a conclusion that the objections raised by the skeptics of twin based heritability must have been correct that twin based heritability overestimates actual heritability.

    Though, as I wrote in one of the earlier comments, the polygenic sore does not have to be linear and additive. Nonlinearity, afaik, was not tested though Hsu wrote one paper last year where he proposed to use 2nd order parabolic nonlinearity but, afaik, he did not implement it. I think the reason he wanted to do it was because his height linear predictor function could not reach twin based heritability even though he used 10,000 SNPs.

    I do not understand why people who make the usual objections to DNA studies of intelligence do not bring the issue of 1,000,000 SNPs used in this paper. For God's sake there is not much more left in coding or nearby regions and all they could get was 11-13%. One reason is that the issue of 1,000,000 SNPs is well hidden in FAQ part and does not appear in the main body of the paper. The abstract of the paper should have been rewritten: 1,271 lead SNPs explain 3.9% of variance. The explained variance increases to 11-13% by adding extra 1,000,000 SNPs into the polygenic score. to avoid gross misrepresentation of the results.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  34. hyperbola says:
    @puropedo
    Very good comment Hyperbola , psychologists , biologists , sociologists , pedagogues , physicians ,etc... are selling snake oil . They are selling a very expensive ,and materialistic " science " , which of course is a very good bussiness for them , but not so much for the general population who pays their academic delusions .

    There is good science and fraudulent science. Unfortunately “our” society now provides vastly more motivation to fraudulent scientists than to honest scientists. A related problem is the use of “pseudo-science” to sell “products”.

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  35. JackOH says:
    @pyrrhus
    Fascinating post, thank you Dr. Thompson! One comment about methodology....
    The correlation between pure intelligence and educational attainment must be much better in Europe than it is in the US...These days in America, virtually everyone, including students with below average IQs, can and does take out student loans and enroll in some kind of open admission, often for-profit, institution of higher "learning." Unfortunately, this generally results in significant debt for no economic benefit, and a level of educational "attainment" unconnected to IQ...We also have lots of scholarship athletes, many of them African-American, who graduate with a 4 year degree but are functionally illiterate.
    On the other end, I have known or taught very bright students who dropped out of college, and even high school, generally to work full time in the IT field or gaming.

    ” . . . [A] level of educational “attainment” [in the States] unconnected to IQ . . .”.

    Exactly so, Pyrrhus. I’m a close observer of my local less selective state university, and it’s my well-informed guess that 80% of the students derive no benefit from attendance, and, what’s worse, there may be actual damage done to these students. They’ll have degrees conferred upon them that will mislead many into believing they’re much more able than they actually are.

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  36. Okechukwu says:
    @APilgrim
    About 'our rootstock' ...

    Neanderthal genes, in Northern Hemisphere populations are directly tied to IQ. Caucasians & Asians have average IQs of over 100. Black Africans & Australian Aboriginals have average IQs below 70. Native Middle Eastern populations and American 'coloreds' have average IQs of approximately 85.

    Neanderthal brains were on average larger than those of fully modern humans, and research points toward symbolic thought.

    Neanderthal genes, in Northern Hemisphere populations are directly tied to IQ.

    Neanderthal was a glorified gorilla. Neanderthal admixture is a minus, not a plus.

    Caucasians & Asians have average IQs of over 100. Black Africans & Australian Aboriginals have average IQs below 70.

    Australian Aborigines carry Neanderthal DNA.

    Caucasians include MENA and Indian sub-continent people. They do not have an average IQ over 100. In fact black Americans have higher IQ’s, as do a large portion black Africans. That’s assuming tha any of this IQ gobbledygook is even real. I think it’s safe to say that IQ is one of the biggest hoaxes ever foisted upon humanity.

    By the way, if Africans have an IQ below 70, which would qualify them for the Special Olympics, then you’d better come up with a plausible explanation as to why I have yet to encounter one as dumb as you are.

    Neanderthal brains were on average larger than those of fully modern humans, and research points toward symbolic thought.

    Research doesn’t have to point toward symbolic thought among early humans. It’s a given.

    Read More
    • Replies: @APilgrim
    Lynn and Vanhanen: Maroon is below 65, Purple is over 105: Deal with it.
    . https://upload.wikimedia.org/wikipedia/commons/5/5e/National_IQ_per_country_-_estimates_by_Lynn_and_Vanhanen_2006.png
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  37. Okechukwu says:
    @res
    Do you think if you repeat this enough times in enough places it will magically become true?

    Regarding 1. Do you understand how additive behavior works?

    Regarding 2. Do you believe the same is true of height?

    Regarding 2. Do you believe the same is true of height?

    Leave it to resident IQ-obsessed troll res to liken something as diatarily regulated as height to the most complex trait we know — a trait that science still doesn’t understand and maybe never will.

    The human brain is literally as opaque and as indecipherable as the universe. On the other hand, eat your vegetables and drink your milk and you’ll grow taller.

    Read More
    • LOL: res
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  38. Okechukwu says:
    @res
    Some of the commenters here might find Eric Turkheimer's take on EA3 more to their liking: http://www.geneticshumanagency.org/gha/ea3-better-formatting/

    Except, no comments allowed ; )

    I was especially impressed by Turkheimer's exceptional modesty:

    I should add that it is first authored by GHA participant James Lee, so I take most of the credit for it.
     
    I guess we are just about to the phase where everyone says "that was obvious" despite having denied exactly that (e.g. hate for The Bell Curve) for decades:

    On the other hand, it must be remembered that the basic theoretical insight that genes have a role to play in human differences like intelligence and education has been well-established for decades.
     
    One good thing is he linked to Robert Plomin's forthcoming book: https://mitpress.mit.edu/books/blueprint

    I guess we are just about to the phase where everyone says “that was obvious” despite having denied exactly that (e.g. hate for The Bell Curve) for decades:

    How misleading and deceptive. Nothing in the paper you linked is an endorsement of The Bell Curve. To this very moment, Turkheimer stands behind his criticism of that book. Don’t take my word for it, Email him.

    Charles Murray is once again peddling junk science about race and IQ

    https://www.vox.com/the-big-idea/2017/5/18/15655638/charles-murray-race-iq-sam-harris-science-free-speech

    On the other hand, it must be remembered that the basic theoretical insight that genes have a role to play in human differences like intelligence and education has been well-established for decades.

    How exactly is this controversial? And how does it go to support The Bell Curve? Notice Turkheimer’s reference is to human (i.e., individual differences) and not group differences. His twins citation could be twins of any race or color. If someone grows up in my extended family they’re probably going to be high achievers educationally because that’s just what we do.

    Read More
    • Replies: @res
    The IQ/genetics denial takes various forms (e.g. related to both individual differences and racial differences). My TBC reference was only an example.

    I saw that Turkheimer (plus Harden and Nisbett) Vox article back when it was first published. Helped me decide those three all have to be read very closely to decipher what they are and are not saying.

    In response to that article, let's take a look at this laughable strawman they give of Murray's nuanced opinion in TBC:

    Murray takes the heritability of intelligence as evidence that it is an essential inborn quality, passed in the genes from parents to children with little modification by environmental factors.
     
    Let's compare that to the paragraph from TBC which I think best captures its position on the role of genetics in racial differences in IQ:

    It the reader is now convinced that either the genetic or environmental explanation has won out to the exclusion of the other, we have not done a sufficiently good job of presenting one side or the other. It seems highly likely to us that both genes and the environment have something to do with racial differences. What might the mix be? We are resolutely agnostic on that issue; as far as we can determine, the evidence does not yet justify an estimate.
     
    See the difference?

    So you are saying there is exactly zero genetic contribution to the Black/White IQ difference in the US? (or are you still maintaining that Blacks actually have a higher genetic IQ but the environment more than compensates?)

    It's good to know these things so I can enjoy a hearty LOL at your expense when you are proven wrong. That you don't think the SNP frequency differences seen for IQ SNPs found so far are convincing evidence there is likely to be some genetic contribution to racial IQ differences says a great deal about your understanding in this area and ability to evaluate the evidence already at hand.

    P.S. The best part about the fallout from that Vox article was the clarification from the authors which made clear that they don't agree on what is true. Steve Sailer talks about that here: http://www.unz.com/isteve/vox-demonizing-charles-murray-is-really-about-protecting-jews/

    If you want a more detailed takedown of the Vox article, this is pretty good: https://medium.com/@houstoneuler/the-cherry-picked-science-in-voxs-charles-murray-article-bd534a9c4476

    Another: https://quillette.com/2017/06/02/getting-voxed-charles-murray-ideology-science-iq/

    And still another: http://quillette.com/2017/06/21/vox-goes-junk-no-good-thats-bit-intelligent-progress/
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  39. utu says:
    @hyperbola
    There may be means to test whether any particular SNP is causal. For example, for rare diseases (where only one or a small handful of genes are involved) there are now attempts to verify and "fix" the medical problem by altering the gene. This type of approach probably becomes too onerous and uninteresting when hundreds of genes are involved.

    By the way, from medicine we know that the vast majority of "disease" conditions involve coding regions of the genome. This is why I prefer to speak about genes rather than SNPs. It also potentially reduces the size of the problem to about 19,000 genes rather than millions of SNPs.

    This is why I prefer to speak about genes rather than SNPs.

    If this is true “that over 98% of the human genome does not encode protein sequences” (wiki) then out of 10 million SNPs only 200,000 belong to the coding regions. In this paper (See my comment #16) to obtain 11-13% variance explanation they effectively used 1 million of SNPs while with 1,271 “lead” SNPs they explained only 3.9%. This means that 800,000 SNPs they used in their polygenic score were form the noncoding regions. Not surprisingly commenter res wrote in his comment #6:

    Supplementary Table 10 has a list of the causal SNPs. One thing that surprises me is how few of them are coding SNPs.

    (BTW, where the Supplementary material is available?)

    It is possibly that the dichotomy of coding/non-coding DNA is false to some extent and there is much more going on than what was assumed so far in genetics. But if it was true that only SNP’s from the coding regions should suffice to explain any trait, then if all 200,000 SNPs from coding regions were already included in the polygenic score and they still could not replicate the twin based heritability of some trait, then the case should be closed with a conclusion that the objections raised by the skeptics of twin based heritability must have been correct that twin based heritability overestimates actual heritability.

    Though, as I wrote in one of the earlier comments, the polygenic sore does not have to be linear and additive. Nonlinearity, afaik, was not tested though Hsu wrote one paper last year where he proposed to use 2nd order parabolic nonlinearity but, afaik, he did not implement it. I think the reason he wanted to do it was because his height linear predictor function could not reach twin based heritability even though he used 10,000 SNPs.

    I do not understand why people who make the usual objections to DNA studies of intelligence do not bring the issue of 1,000,000 SNPs used in this paper. For God’s sake there is not much more left in coding or nearby regions and all they could get was 11-13%. One reason is that the issue of 1,000,000 SNPs is well hidden in FAQ part and does not appear in the main body of the paper. The abstract of the paper should have been rewritten: 1,271 lead SNPs explain 3.9% of variance. The explained variance increases to 11-13% by adding extra 1,000,000 SNPs into the polygenic score. to avoid gross misrepresentation of the results.

    Read More
    • Replies: @res

    (BTW, where the Supplementary material is available?)
     
    https://www.nature.com/articles/s41588-018-0147-3#Sec34

    Though, as I wrote in one of the earlier comments, the polygenic sore does not have to be linear and additive. Nonlinearity, afaik, was not tested though Hsu wrote one paper last year where he proposed to use 2nd order parabolic nonlinearity but, afaik, he did not implement it. I think the reason he wanted to do it was because his height linear predictor function could not reach twin based heritability even though he used 10,000 SNPs.
     
    I think the reason is inadequate sample size. To the best of my knowledge they have not published nonlinear prediction of height even though the UKBB sample size appears to be have been sufficient to "solve" the additive heritability of height.

    The contention is not that nonlinearity does not matter. Rather that most of the heritability is accounted for by the linear effects.

    I would be much more comfortable interpreting the 11-13% variance explained number from this paper if I knew how much of the variance was explained by their baseline model.
    , @hyperbola
    (1) There may still be further effects associated with non-coding regions. Chromatin structure and its packaging in the nucleus is one obvious area for such effects. There are some studies that attempt to look at changes in the spatial organization of nuclear chromatin as a function of environmental conditions for a cell. It includes things like changes in the interaction of chromatin with the nuclear envelope (hence potential changes in expression of different regions of chromatin).

    (2) Many SNPs may be functionally neutral, i.e. any effect they produce is to small to be observed.


    These studies and their "canned" forms of analysis are wearing out my patience to keep reading them!
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  40. Tim too says:

    Expression levels of various trophic factors (typically called growth factors of some kind, like BDNF, IGF, etc) are not exclusively determined by genetics. Epigenetics and diet among other things, like stress and toxic exposures, chronic or acute, affect expression levels, of trophic factors crucial for intelligence, such as the neurotrophic factors like BDNF. Thus the usage of genetic data, including SNPs is at best supportive, or permissive of intelligence, rather than determinative. Genetics may set some sort of upper level or bound for intelligence, but not set a lower level. (Tissue damage runs the other direction.)

    This is aside from environmental factors such as family culture and other cultural factors.

    This issue is too difficult to pin down to purely genetic factors, because of the large environmental and epigenetic effects.

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter Display All Comments
  41. res says:
    @Okechukwu

    I guess we are just about to the phase where everyone says “that was obvious” despite having denied exactly that (e.g. hate for The Bell Curve) for decades:
     
    How misleading and deceptive. Nothing in the paper you linked is an endorsement of The Bell Curve. To this very moment, Turkheimer stands behind his criticism of that book. Don't take my word for it, Email him.

    Charles Murray is once again peddling junk science about race and IQ

    https://www.vox.com/the-big-idea/2017/5/18/15655638/charles-murray-race-iq-sam-harris-science-free-speech


    On the other hand, it must be remembered that the basic theoretical insight that genes have a role to play in human differences like intelligence and education has been well-established for decades.
     
    How exactly is this controversial? And how does it go to support The Bell Curve? Notice Turkheimer's reference is to human (i.e., individual differences) and not group differences. His twins citation could be twins of any race or color. If someone grows up in my extended family they're probably going to be high achievers educationally because that's just what we do.

    The IQ/genetics denial takes various forms (e.g. related to both individual differences and racial differences). My TBC reference was only an example.

    I saw that Turkheimer (plus Harden and Nisbett) Vox article back when it was first published. Helped me decide those three all have to be read very closely to decipher what they are and are not saying.

    In response to that article, let’s take a look at this laughable strawman they give of Murray’s nuanced opinion in TBC:

    Murray takes the heritability of intelligence as evidence that it is an essential inborn quality, passed in the genes from parents to children with little modification by environmental factors.

    Let’s compare that to the paragraph from TBC which I think best captures its position on the role of genetics in racial differences in IQ:

    It the reader is now convinced that either the genetic or environmental explanation has won out to the exclusion of the other, we have not done a sufficiently good job of presenting one side or the other. It seems highly likely to us that both genes and the environment have something to do with racial differences. What might the mix be? We are resolutely agnostic on that issue; as far as we can determine, the evidence does not yet justify an estimate.

    See the difference?

    So you are saying there is exactly zero genetic contribution to the Black/White IQ difference in the US? (or are you still maintaining that Blacks actually have a higher genetic IQ but the environment more than compensates?)

    It’s good to know these things so I can enjoy a hearty LOL at your expense when you are proven wrong. That you don’t think the SNP frequency differences seen for IQ SNPs found so far are convincing evidence there is likely to be some genetic contribution to racial IQ differences says a great deal about your understanding in this area and ability to evaluate the evidence already at hand.

    P.S. The best part about the fallout from that Vox article was the clarification from the authors which made clear that they don’t agree on what is true. Steve Sailer talks about that here: http://www.unz.com/isteve/vox-demonizing-charles-murray-is-really-about-protecting-jews/

    If you want a more detailed takedown of the Vox article, this is pretty good: https://medium.com/@houstoneuler/the-cherry-picked-science-in-voxs-charles-murray-article-bd534a9c4476

    Another: https://quillette.com/2017/06/02/getting-voxed-charles-murray-ideology-science-iq/

    And still another: http://quillette.com/2017/06/21/vox-goes-junk-no-good-thats-bit-intelligent-progress/

    Read More
    • Replies: @Okechukwu

    The IQ/genetics denial takes various forms (e.g. related to both individual differences and racial differences). My TBC reference was only an example.
     
    No one denies individual differences in intelligence. This is strawmanning at its best (or worst).

    So you are saying there is exactly zero genetic contribution to the Black/White IQ difference in the US? (or are you still maintaining that Blacks actually have a higher genetic IQ but the environment more than compensates?)
     
    Given the plasticity of IQ and given the tremendous within race variation and between race overlap, you'd have to be pretty illiterate scientifically to claim that genes play a role in the black/white IQ gap. Moreover, a genetic interpretation wouldn't necessarily support the conclusion you are pining for. If you followed the debate that was instigated by the Vox article, you'll recall that Ezra Klein called James Flynn a few days before his encounter with Sam Harris. Flynn went on to say that it is entirely possible that the 10-point IQ gap reflects a 12 point environmental difference and a negative 2 point genetic difference. That is, blacks could very well have an inherent advantage in intelligence that is overwhelmed by environmental factors. This to me is more plausible than the alternative.

    It’s good to know these things so I can enjoy a hearty LOL at your expense when you are proven wrong. That you don’t think the SNP frequency differences seen for IQ SNPs found so far are convincing evidence there is likely to be some genetic contribution to racial IQ differences says a great deal about your understanding in this area and ability to evaluate the evidence already at hand.
     
    This is absurd. Nothing has been seen except in the pseudoscience circles of the dark web. There's nothing you can present that can't be picked apart and comprehensively rebutted. This is precisely why your "data" is trafficked only within the pseudoscience circles of the dark web. You IQists may be dumb, but I will give you credit for your competence in propaganda, misdirection and your exceptional skills in hiding away from the disinfecting light of real scientific inquiry. Take it to a real scientific conference, and then explain why such and such black genius is able to exist while lacking the alleged "IQ-related SNP frequencies" you claim is discretely differentiated between blacks and whites. You're holding a house of cards, bruh. Even a baby's breath can blown it down.

    P.S. The best part about the fallout from that Vox article was the clarification from the authors which made clear that they don’t agree on what is true. Steve Sailer talks about that here: http://www.unz.com/isteve/vox-demonizing-charles-murray-is-really-about-protecting-jews/
     
    Of course that's how you and Sailer would read the follow-up article. To a rational, unbiased observer the second article served only to further bury Murray and his racist pseudoscience with much more detail than the first.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  42. res says:
    @utu

    This is why I prefer to speak about genes rather than SNPs.
     
    If this is true "that over 98% of the human genome does not encode protein sequences" (wiki) then out of 10 million SNPs only 200,000 belong to the coding regions. In this paper (See my comment #16) to obtain 11-13% variance explanation they effectively used 1 million of SNPs while with 1,271 "lead" SNPs they explained only 3.9%. This means that 800,000 SNPs they used in their polygenic score were form the noncoding regions. Not surprisingly commenter res wrote in his comment #6:

    Supplementary Table 10 has a list of the causal SNPs. One thing that surprises me is how few of them are coding SNPs.
     
    (BTW, where the Supplementary material is available?)

    It is possibly that the dichotomy of coding/non-coding DNA is false to some extent and there is much more going on than what was assumed so far in genetics. But if it was true that only SNP's from the coding regions should suffice to explain any trait, then if all 200,000 SNPs from coding regions were already included in the polygenic score and they still could not replicate the twin based heritability of some trait, then the case should be closed with a conclusion that the objections raised by the skeptics of twin based heritability must have been correct that twin based heritability overestimates actual heritability.

    Though, as I wrote in one of the earlier comments, the polygenic sore does not have to be linear and additive. Nonlinearity, afaik, was not tested though Hsu wrote one paper last year where he proposed to use 2nd order parabolic nonlinearity but, afaik, he did not implement it. I think the reason he wanted to do it was because his height linear predictor function could not reach twin based heritability even though he used 10,000 SNPs.

    I do not understand why people who make the usual objections to DNA studies of intelligence do not bring the issue of 1,000,000 SNPs used in this paper. For God's sake there is not much more left in coding or nearby regions and all they could get was 11-13%. One reason is that the issue of 1,000,000 SNPs is well hidden in FAQ part and does not appear in the main body of the paper. The abstract of the paper should have been rewritten: 1,271 lead SNPs explain 3.9% of variance. The explained variance increases to 11-13% by adding extra 1,000,000 SNPs into the polygenic score. to avoid gross misrepresentation of the results.

    (BTW, where the Supplementary material is available?)

    https://www.nature.com/articles/s41588-018-0147-3#Sec34

    Though, as I wrote in one of the earlier comments, the polygenic sore does not have to be linear and additive. Nonlinearity, afaik, was not tested though Hsu wrote one paper last year where he proposed to use 2nd order parabolic nonlinearity but, afaik, he did not implement it. I think the reason he wanted to do it was because his height linear predictor function could not reach twin based heritability even though he used 10,000 SNPs.

    I think the reason is inadequate sample size. To the best of my knowledge they have not published nonlinear prediction of height even though the UKBB sample size appears to be have been sufficient to “solve” the additive heritability of height.

    The contention is not that nonlinearity does not matter. Rather that most of the heritability is accounted for by the linear effects.

    I would be much more comfortable interpreting the 11-13% variance explained number from this paper if I knew how much of the variance was explained by their baseline model.

    Read More
    • Replies: @utu
    Thanks for the link.

    I think the reason is inadequate sample size.
     
    I do not see how a larger sample would helped him? He pre-filterd SNPs down to 50k of them and then as he was looking for solution with larger and larger number of SNPs he hit a plateau where by adding more SNP's R^2 on the validation sample was no longer increasing. A larger sample would not change it. But by having a nonlinear polygenic score it is possible that more could have been squeezed out from the data.

    A larger sample could change anything only if it affected his pre-filtering that would lead to a different 50k subset of SNPs. I do not see it very likely. He based his pre-filtering on individual correlation of SNP with the trait. Would different SNPs ended up in his pre-filtered subset?
    , @utu

    I would be much more comfortable interpreting the 11-13% variance explained number from this paper if I knew how much of the variance was explained by their baseline model.
     
    We talked about it. It is possible that the baseline model removes too much and w/o it or with a more gentle one the explained incremental variance would be larger. When Turkheimer writes "The authors of the study are cautious and thoughtful in their interpretation of the results, far more so than the people above." he does not, I think, address this issue but indicates that the authors are conservative. But perhaps they are conservative and rather report lower than higher variance.

    Still you have not touched the issue of 1,000,000 SNPs in the polygenic score and why it is not discussed in the main body of the paper.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  43. utu says:
    @res

    (BTW, where the Supplementary material is available?)
     
    https://www.nature.com/articles/s41588-018-0147-3#Sec34

    Though, as I wrote in one of the earlier comments, the polygenic sore does not have to be linear and additive. Nonlinearity, afaik, was not tested though Hsu wrote one paper last year where he proposed to use 2nd order parabolic nonlinearity but, afaik, he did not implement it. I think the reason he wanted to do it was because his height linear predictor function could not reach twin based heritability even though he used 10,000 SNPs.
     
    I think the reason is inadequate sample size. To the best of my knowledge they have not published nonlinear prediction of height even though the UKBB sample size appears to be have been sufficient to "solve" the additive heritability of height.

    The contention is not that nonlinearity does not matter. Rather that most of the heritability is accounted for by the linear effects.

    I would be much more comfortable interpreting the 11-13% variance explained number from this paper if I knew how much of the variance was explained by their baseline model.

    Thanks for the link.

    I think the reason is inadequate sample size.

    I do not see how a larger sample would helped him? He pre-filterd SNPs down to 50k of them and then as he was looking for solution with larger and larger number of SNPs he hit a plateau where by adding more SNP’s R^2 on the validation sample was no longer increasing. A larger sample would not change it. But by having a nonlinear polygenic score it is possible that more could have been squeezed out from the data.

    A larger sample could change anything only if it affected his pre-filtering that would lead to a different 50k subset of SNPs. I do not see it very likely. He based his pre-filtering on individual correlation of SNP with the trait. Would different SNPs ended up in his pre-filtered subset?

    Read More
    • Replies: @res
    Adding nonlinear terms either doubles (including square terms) or squares (!, including interaction terms) the number of variables. I am pretty sure that has implications for the sample size required.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  44. Okechukwu says:
    @res
    The IQ/genetics denial takes various forms (e.g. related to both individual differences and racial differences). My TBC reference was only an example.

    I saw that Turkheimer (plus Harden and Nisbett) Vox article back when it was first published. Helped me decide those three all have to be read very closely to decipher what they are and are not saying.

    In response to that article, let's take a look at this laughable strawman they give of Murray's nuanced opinion in TBC:

    Murray takes the heritability of intelligence as evidence that it is an essential inborn quality, passed in the genes from parents to children with little modification by environmental factors.
     
    Let's compare that to the paragraph from TBC which I think best captures its position on the role of genetics in racial differences in IQ:

    It the reader is now convinced that either the genetic or environmental explanation has won out to the exclusion of the other, we have not done a sufficiently good job of presenting one side or the other. It seems highly likely to us that both genes and the environment have something to do with racial differences. What might the mix be? We are resolutely agnostic on that issue; as far as we can determine, the evidence does not yet justify an estimate.
     
    See the difference?

    So you are saying there is exactly zero genetic contribution to the Black/White IQ difference in the US? (or are you still maintaining that Blacks actually have a higher genetic IQ but the environment more than compensates?)

    It's good to know these things so I can enjoy a hearty LOL at your expense when you are proven wrong. That you don't think the SNP frequency differences seen for IQ SNPs found so far are convincing evidence there is likely to be some genetic contribution to racial IQ differences says a great deal about your understanding in this area and ability to evaluate the evidence already at hand.

    P.S. The best part about the fallout from that Vox article was the clarification from the authors which made clear that they don't agree on what is true. Steve Sailer talks about that here: http://www.unz.com/isteve/vox-demonizing-charles-murray-is-really-about-protecting-jews/

    If you want a more detailed takedown of the Vox article, this is pretty good: https://medium.com/@houstoneuler/the-cherry-picked-science-in-voxs-charles-murray-article-bd534a9c4476

    Another: https://quillette.com/2017/06/02/getting-voxed-charles-murray-ideology-science-iq/

    And still another: http://quillette.com/2017/06/21/vox-goes-junk-no-good-thats-bit-intelligent-progress/

    The IQ/genetics denial takes various forms (e.g. related to both individual differences and racial differences). My TBC reference was only an example.

    No one denies individual differences in intelligence. This is strawmanning at its best (or worst).

    So you are saying there is exactly zero genetic contribution to the Black/White IQ difference in the US? (or are you still maintaining that Blacks actually have a higher genetic IQ but the environment more than compensates?)

    Given the plasticity of IQ and given the tremendous within race variation and between race overlap, you’d have to be pretty illiterate scientifically to claim that genes play a role in the black/white IQ gap. Moreover, a genetic interpretation wouldn’t necessarily support the conclusion you are pining for. If you followed the debate that was instigated by the Vox article, you’ll recall that Ezra Klein called James Flynn a few days before his encounter with Sam Harris. Flynn went on to say that it is entirely possible that the 10-point IQ gap reflects a 12 point environmental difference and a negative 2 point genetic difference. That is, blacks could very well have an inherent advantage in intelligence that is overwhelmed by environmental factors. This to me is more plausible than the alternative.

    It’s good to know these things so I can enjoy a hearty LOL at your expense when you are proven wrong. That you don’t think the SNP frequency differences seen for IQ SNPs found so far are convincing evidence there is likely to be some genetic contribution to racial IQ differences says a great deal about your understanding in this area and ability to evaluate the evidence already at hand.

    This is absurd. Nothing has been seen except in the pseudoscience circles of the dark web. There’s nothing you can present that can’t be picked apart and comprehensively rebutted. This is precisely why your “data” is trafficked only within the pseudoscience circles of the dark web. You IQists may be dumb, but I will give you credit for your competence in propaganda, misdirection and your exceptional skills in hiding away from the disinfecting light of real scientific inquiry. Take it to a real scientific conference, and then explain why such and such black genius is able to exist while lacking the alleged “IQ-related SNP frequencies” you claim is discretely differentiated between blacks and whites. You’re holding a house of cards, bruh. Even a baby’s breath can blown it down.

    P.S. The best part about the fallout from that Vox article was the clarification from the authors which made clear that they don’t agree on what is true. Steve Sailer talks about that here: http://www.unz.com/isteve/vox-demonizing-charles-murray-is-really-about-protecting-jews/

    Of course that’s how you and Sailer would read the follow-up article. To a rational, unbiased observer the second article served only to further bury Murray and his racist pseudoscience with much more detail than the first.

    Read More
    • Replies: @res

    No one denies individual differences in intelligence. This is strawmanning at its best (or worst).
     
    I specifically referred to denying the influence of genetics on intelligence. In other words, you strawmanned me by leaving out the "genetics" qualifier. But you are right that your first sentence was an example of strawmanning at its worst (that was what you meant, right?).

    As for no one denying the influence of genetics of on intelligence, if that is so then why did someone feel the need to write this article in 2017? Yes, There Is a Genetic Component to Intelligence https://www.thecut.com/2017/05/genetics-intelligence.html


    Given the plasticity of IQ and given the tremendous within race variation and between race overlap, you’d have to be pretty illiterate scientifically to claim that genes play a role in the black/white IQ gap. Moreover, a genetic interpretation wouldn’t necessarily support the conclusion you are pining for. If you followed the debate that was instigated by the Vox article, you’ll recall that Ezra Klein called James Flynn a few days before his encounter with Sam Harris. Flynn went on to say that it is entirely possible that the 10-point IQ gap reflects a 12 point environmental difference and a negative 2 point genetic difference. That is, blacks could very well have an inherent advantage in intelligence that is overwhelmed by environmental factors. This to me is more plausible than the alternative.
     
    That was some Turkheimer worthy obfuscation. Let's make this simple. Yes or no, do you believe there is exactly zero genetic contribution to the Black/White IQ difference in the US? Any answer other than an explicit "yes" indicates agreement with the conclusion I quoted from The Bell Curve. To repeat the most important part: "It seems highly likely to us that both genes and the environment have something to do with racial differences."

    On what basis do you think a black genetic advantage in intelligence is more likely than the opposite? I believe the opposite is more likely (has higher probability) because the required environmental difference is smaller if the phenotypic difference has the same sign as the genetic difference. A smaller required environmental difference is more likely.


    There’s nothing you can present that can’t be picked apart and comprehensively rebutted.
     
    The SNP frequencies differ between the races for the first (and likely almost all) IQ SNPs found. That is reality. Sorry you don't like it.

    You’re holding a house of cards, bruh. Even a baby’s breath can blown it down.

     

    Now that was some world class argumentation. I tremble before the awesomeness of your debating skills. /sarc

    To a rational, unbiased observer the second article served only to further bury Murray and his racist pseudoscience with much more detail than the first.
     
    Because the failure of the three authors to agree on a single response demonstrates just how compelling their arguments are. LOL!
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  45. utu says:
    @res

    (BTW, where the Supplementary material is available?)
     
    https://www.nature.com/articles/s41588-018-0147-3#Sec34

    Though, as I wrote in one of the earlier comments, the polygenic sore does not have to be linear and additive. Nonlinearity, afaik, was not tested though Hsu wrote one paper last year where he proposed to use 2nd order parabolic nonlinearity but, afaik, he did not implement it. I think the reason he wanted to do it was because his height linear predictor function could not reach twin based heritability even though he used 10,000 SNPs.
     
    I think the reason is inadequate sample size. To the best of my knowledge they have not published nonlinear prediction of height even though the UKBB sample size appears to be have been sufficient to "solve" the additive heritability of height.

    The contention is not that nonlinearity does not matter. Rather that most of the heritability is accounted for by the linear effects.

    I would be much more comfortable interpreting the 11-13% variance explained number from this paper if I knew how much of the variance was explained by their baseline model.

    I would be much more comfortable interpreting the 11-13% variance explained number from this paper if I knew how much of the variance was explained by their baseline model.

    We talked about it. It is possible that the baseline model removes too much and w/o it or with a more gentle one the explained incremental variance would be larger. When Turkheimer writes “The authors of the study are cautious and thoughtful in their interpretation of the results, far more so than the people above.” he does not, I think, address this issue but indicates that the authors are conservative. But perhaps they are conservative and rather report lower than higher variance.

    Still you have not touched the issue of 1,000,000 SNPs in the polygenic score and why it is not discussed in the main body of the paper.

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  46. res says:
    @utu
    Thanks for the link.

    I think the reason is inadequate sample size.
     
    I do not see how a larger sample would helped him? He pre-filterd SNPs down to 50k of them and then as he was looking for solution with larger and larger number of SNPs he hit a plateau where by adding more SNP's R^2 on the validation sample was no longer increasing. A larger sample would not change it. But by having a nonlinear polygenic score it is possible that more could have been squeezed out from the data.

    A larger sample could change anything only if it affected his pre-filtering that would lead to a different 50k subset of SNPs. I do not see it very likely. He based his pre-filtering on individual correlation of SNP with the trait. Would different SNPs ended up in his pre-filtered subset?

    Adding nonlinear terms either doubles (including square terms) or squares (!, including interaction terms) the number of variables. I am pretty sure that has implications for the sample size required.

    Read More
    • Agree: utu
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  47. res says:
    @Okechukwu

    The IQ/genetics denial takes various forms (e.g. related to both individual differences and racial differences). My TBC reference was only an example.
     
    No one denies individual differences in intelligence. This is strawmanning at its best (or worst).

    So you are saying there is exactly zero genetic contribution to the Black/White IQ difference in the US? (or are you still maintaining that Blacks actually have a higher genetic IQ but the environment more than compensates?)
     
    Given the plasticity of IQ and given the tremendous within race variation and between race overlap, you'd have to be pretty illiterate scientifically to claim that genes play a role in the black/white IQ gap. Moreover, a genetic interpretation wouldn't necessarily support the conclusion you are pining for. If you followed the debate that was instigated by the Vox article, you'll recall that Ezra Klein called James Flynn a few days before his encounter with Sam Harris. Flynn went on to say that it is entirely possible that the 10-point IQ gap reflects a 12 point environmental difference and a negative 2 point genetic difference. That is, blacks could very well have an inherent advantage in intelligence that is overwhelmed by environmental factors. This to me is more plausible than the alternative.

    It’s good to know these things so I can enjoy a hearty LOL at your expense when you are proven wrong. That you don’t think the SNP frequency differences seen for IQ SNPs found so far are convincing evidence there is likely to be some genetic contribution to racial IQ differences says a great deal about your understanding in this area and ability to evaluate the evidence already at hand.
     
    This is absurd. Nothing has been seen except in the pseudoscience circles of the dark web. There's nothing you can present that can't be picked apart and comprehensively rebutted. This is precisely why your "data" is trafficked only within the pseudoscience circles of the dark web. You IQists may be dumb, but I will give you credit for your competence in propaganda, misdirection and your exceptional skills in hiding away from the disinfecting light of real scientific inquiry. Take it to a real scientific conference, and then explain why such and such black genius is able to exist while lacking the alleged "IQ-related SNP frequencies" you claim is discretely differentiated between blacks and whites. You're holding a house of cards, bruh. Even a baby's breath can blown it down.

    P.S. The best part about the fallout from that Vox article was the clarification from the authors which made clear that they don’t agree on what is true. Steve Sailer talks about that here: http://www.unz.com/isteve/vox-demonizing-charles-murray-is-really-about-protecting-jews/
     
    Of course that's how you and Sailer would read the follow-up article. To a rational, unbiased observer the second article served only to further bury Murray and his racist pseudoscience with much more detail than the first.

    No one denies individual differences in intelligence. This is strawmanning at its best (or worst).

    I specifically referred to denying the influence of genetics on intelligence. In other words, you strawmanned me by leaving out the “genetics” qualifier. But you are right that your first sentence was an example of strawmanning at its worst (that was what you meant, right?).

    As for no one denying the influence of genetics of on intelligence, if that is so then why did someone feel the need to write this article in 2017? Yes, There Is a Genetic Component to Intelligence https://www.thecut.com/2017/05/genetics-intelligence.html

    Given the plasticity of IQ and given the tremendous within race variation and between race overlap, you’d have to be pretty illiterate scientifically to claim that genes play a role in the black/white IQ gap. Moreover, a genetic interpretation wouldn’t necessarily support the conclusion you are pining for. If you followed the debate that was instigated by the Vox article, you’ll recall that Ezra Klein called James Flynn a few days before his encounter with Sam Harris. Flynn went on to say that it is entirely possible that the 10-point IQ gap reflects a 12 point environmental difference and a negative 2 point genetic difference. That is, blacks could very well have an inherent advantage in intelligence that is overwhelmed by environmental factors. This to me is more plausible than the alternative.

    That was some Turkheimer worthy obfuscation. Let’s make this simple. Yes or no, do you believe there is exactly zero genetic contribution to the Black/White IQ difference in the US? Any answer other than an explicit “yes” indicates agreement with the conclusion I quoted from The Bell Curve. To repeat the most important part: “It seems highly likely to us that both genes and the environment have something to do with racial differences.”

    On what basis do you think a black genetic advantage in intelligence is more likely than the opposite? I believe the opposite is more likely (has higher probability) because the required environmental difference is smaller if the phenotypic difference has the same sign as the genetic difference. A smaller required environmental difference is more likely.

    There’s nothing you can present that can’t be picked apart and comprehensively rebutted.

    The SNP frequencies differ between the races for the first (and likely almost all) IQ SNPs found. That is reality. Sorry you don’t like it.

    You’re holding a house of cards, bruh. Even a baby’s breath can blown it down.

    Now that was some world class argumentation. I tremble before the awesomeness of your debating skills. /sarc

    To a rational, unbiased observer the second article served only to further bury Murray and his racist pseudoscience with much more detail than the first.

    Because the failure of the three authors to agree on a single response demonstrates just how compelling their arguments are. LOL!

    Read More
    • Replies: @Okechukwu

    I specifically referred to denying the influence of genetics on intelligence.
     
    No one has ever denied genetic or inherited intelligence. Even prehistoric humans knew that attributes like intelligence can be transmitted from parent to offspring.

    Yes or no, do you believe there is exactly zero genetic contribution to the Black/White IQ difference in the US?
     
    It's pointless. Nigerian-Americans are reputed to have an IQ of 110. Is their nearly 1 standard deviation advantage over whites effectuated genetically? There are all kinds of group IQ disparities. Why are you insistent on a genetic cause only with respect to the black/white gap in the US? Is the southern and northern white American IQ gap genetically based? How about the Catholic and Protestant Irish IQ gap? Or the Japanese and ethnic Korean IQ gap in Japan? Etc., Etc., Etc.

    On what basis do you think a black genetic advantage in intelligence is more likely than the opposite?
     
    Because after 400 years of oppression and cultural and social ostracization they are only 9 or 10 points behind. Flynn is too conservative. Whites should be 20-30 points ahead rather than a mere 9.

    The SNP frequencies differ between the races for the first (and likely almost all) IQ SNPs found. That is reality. Sorry you don’t like it.
     
    If you weren't an idiot you would be highly skeptical of that data, given that smart and genius black people do exist ... Duh. Btw, there is no such thing as an "IQ SNP."

    Because the failure of the three authors to agree on a single response demonstrates just how compelling their arguments are. LOL!
     
    That's only your opinion. Those three authors are firmly anchored within the mainstream of scientific thought. You, Sailer and Murray are the outliers. Furthermore, we know that your thoughts and reactions are motivated by varying strains of white nationalism/supremacism. So you are not credible actors. Therefore your conclusions are unreliable.
    , @RaceRealist88
    "the influence of genetics of on intelligence"

    Genes can't influence intelligence because there are no psychophysical or psychological laws.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  48. It amazes me how much fear and loathing the singular issue of human intelligence still engenders. Everyone has too much skin in the game. We are biological computers that are afraid of our own obsolescence. Realize that the singularity oft spoken of in the field of artificial intelligence pertains instead to the biological. The race must have already begun.

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter Display All Comments
  49. APilgrim says:
    @Okechukwu

    Neanderthal genes, in Northern Hemisphere populations are directly tied to IQ.
     
    Neanderthal was a glorified gorilla. Neanderthal admixture is a minus, not a plus.

    Caucasians & Asians have average IQs of over 100. Black Africans & Australian Aboriginals have average IQs below 70.
     
    Australian Aborigines carry Neanderthal DNA.

    Caucasians include MENA and Indian sub-continent people. They do not have an average IQ over 100. In fact black Americans have higher IQ's, as do a large portion black Africans. That's assuming tha any of this IQ gobbledygook is even real. I think it's safe to say that IQ is one of the biggest hoaxes ever foisted upon humanity.

    By the way, if Africans have an IQ below 70, which would qualify them for the Special Olympics, then you'd better come up with a plausible explanation as to why I have yet to encounter one as dumb as you are.


    Neanderthal brains were on average larger than those of fully modern humans, and research points toward symbolic thought.
     
    Research doesn't have to point toward symbolic thought among early humans. It's a given.

    Lynn and Vanhanen: Maroon is below 65, Purple is over 105: Deal with it.
    .

    Read More
    • Replies: @Tim too
    Malnutrition.
    , @Okechukwu

    Lynn and Vanhanen: Maroon is below 65, Purple is over 105: Deal with it.
     
    I'm familiar with the chart. What's shocking is that someone's dumb enough to believe it. Rather ironic in a discussion of intelligence.

    Funny how African athletes are able to compete in the normal Olympics rather than the Special Olympics.

    Mental retardation is cognitive limitation as characterized by scores greater than 2 standard deviations below the mean on a valid intelligence quotient (IQ) measure, with limitation of adaptive function in communication, self‐care, daily living skills at home or in the community, or social skills.

    https://www.sciencedirect.com/topics/neuroscience/mental-retardation

    Not only do Africans not have any of these issues, when they immigrate to the west they outperform whites.

    African immigrants are more educated than most — including people born in U.S.

    Batalova's research found that of the 1.4 million who are 25 and older, 41% have a bachelor's degree, compared with 30% of all immigrants and 32% of the U.S.-born population. Of the 19,000 U.S. immigrants from Norway — a country Trump reportedly told lawmakers is a good source of immigrants — 38% have college educations.

    The New American Economy study found that 1 in 3 of these undergraduate degrees were focused on science, technology, engineering and math — "training heavily in demand by today's employers."

    That report also found that African immigrants were significantly more likely to have graduate degrees. A total of 16% had a master's degree, medical degree, law degree or a doctorate, compared with 11% of the U.S.-born population.

    http://www.latimes.com/world/africa/la-fg-global-african-immigrants-explainer-20180112-story.html

    I don't know who's more stupid, Lynn and Vanhanen or people like you who accept their "research."
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  50. j2 says:
    @hyperbola
    There are any number of medical conditions (rare diseases) where sample sizes much less than one million are adequate to define the genes responsible for the disease. Often one finds that one or a handful of genes that are both necessary and sufficient to cause the disease.

    What the million-sample "IQ" studies show is that there is NO small number of genes that are necessary and sufficient for producing such a complex (and very poorly defined) trait. For "IQ", THE RESULTS ARE IN (unless you wish to propose that an extremely improbable set of people have ended up in your million person sample).

    Waving magic solutions like "polygenic" (why should "additive" be linear - complex interdependence is physiologically much more reasonable) is simply an attempt to cover up for failure. If there were as few as 10 genes, the studies should already have given an overwhelmingly clear answer. Time to acknowledge that for "IQ" the situation is so complex that even with 30 genes essentially all of humanity would have to be measured to find those 30 genes.

    hyperbola writes:
    “(why should “additive” be linear – complex interdependence is physiologically much more reasonable)”

    This is a mathematical issue. Most of these complex interdependences are continuous, differentiable functions. A small effect is then f(p)+f ‘(p)dx, it is linear (i.e., effects are additive). This is true for small effects and it explains some part why such additive constructs like a polygenic score cannot explain much of the effect: it is additive (=linear) only for small changes. In order to find out the whole function (all partial derivatives) would require very much study.

    The mathematically more reasonable way to evaluate the effect of SNPs to IQ would be to show that an SNP has a small effect rather than no effect (as we cannot find out the whole effect without knowing the function) and then scan over all genes. The set of genes that have any effect account for the whole genetic effect, estimated to be 60-80%. Then you can assume that the effects are distributed in some way, like normally (a few have larger effects, most intermediate) and to calculate the average effect.

    Read More
    • Replies: @hyperbola
    The assumption of linearity over small changes (dx) is not warranted by any real evidence - a step function rather than a continuous function might be more characteristic of physiological functions. Imagine 2 proteins that interact directly. Their interaction might be turned off completely by a change that affects either one of them. Whether this interaction is dominant for "our" target "observable" remains an open question.

    Maybe we have to think about this in completely different ways. The reason that we cannot identify the dominant "mutations" for "IQ" is that they are lethal and we never have them included in the samples. The remaining "mutations" observed in our samples have only rather peripheral effects (to avoid lethality), may well be part of highly parallel partly redundant "functions" and will be essentially useless for prediction of things like "IQ".

    Speculation, but maybe it serves to imagine that the "canned" analyses currently being applied to these kinds of problems is far from being definitive.
    , @utu
    Keep in mind that it is not even possible to writes something like this "f(p)+f ‘(p)dx" in the case of the polygenic score. The independent variables are binary. One either have a given allele of SNP or not. The linearity in question concerns multiple variables not a single variable. The nonlinearity in terms of one variable does not even makes sense because variable is basically binary. If x and y are two different SNPs GWAS can approximate the answer to what is the difference in trait when x=0 or 1, i.e., when you have a given allele of the SNP or not. The same can be done for SNP y. But it is harder to answer whether the effect of x and y together is a sum of effects due to separate effects of x and of y. And when you have more and more SNPs that you want to include in the polygenic score it becomes harder and harder to disentangle their non liner interactions. In case of two binary variables x and y you can write polygenic score as: ax+by or ax+by+cxy. In the second case the nonlinear term xy, which kicks in only when both x and y are non zero, takes care of the possibility that the effect of two SNPs is not equal to the sum of the effects. This formulation makes it possible to account for the case when x and y individually have no effect (a=b=0) but in tandem they do. More or less this is how it should be done. The problem is that when you consider n SNPs in a linear case you have n variables but in nonlinear case with mixed terms you have additionally n*(n-1)/2 terms to consider. In the case of 1,000 SNPs you'll end up with 500,000 variables to be determined.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  51. KnowsMore says:

    Regardless of this pseudo-scientific hype by affluenza victims hoping to pull up the ladder, it is plain from the mere existence of brilliant members of disadvantaged minorities, raised in fortunate circumstances, that there is no general genetic disadvantage in race.

    Your measurements only diagnose your science as faulty. You find the error.

    It is amazing what volumes of obfuscation and pseudoscience the racist will stoop to or believe. UNZ should not publish this crap.

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter Display All Comments
  52. Tim too says:
    @APilgrim
    Lynn and Vanhanen: Maroon is below 65, Purple is over 105: Deal with it.
    . https://upload.wikimedia.org/wikipedia/commons/5/5e/National_IQ_per_country_-_estimates_by_Lynn_and_Vanhanen_2006.png

    Malnutrition.

    Read More
    • Replies: @res
    I was initially inclined to dismiss that glib response, but you have made some interesting comments in your short time posting at unz.com so perhaps it is worth engaging.

    Do you think malnutrition is the only factor explaining the sea of red in APilgrim's graphic? Surely there are other countries that are malnourished? Let's take a look: https://ourworldindata.org/hunger-and-undernourishment

    (edit: looks like SVG files do not embed here, click through to see map)

    https://d33wubrfki0l68.cloudfront.net/fdb5a6c43da0d5bc865808e9767c97dddc0fa053/6ea73/exports/global-hunger-index-e3bcb6c247c02f1e57c637d2d9c7fe75_v1_850x600.svg

    Try another:

    http://cdn3.chartsbin.com/chartimages/l_3355_474b042fc1dc129b6e492440c7bd78aa

    The choropleths are somewhat similar, but there is a notable disconnect between the Subcontinent and sub-Saharan Africa. How do you explain that?

    , @APilgrim
    Most of the Purple Areas experienced STARVATION, during most of the 20th Century.

    During WWII, the entire Purple area was malnourished.

    North Koreans are still grossly malnourished. NORKS have TWICE the IQ of Blacks.

    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  53. res says:
    @Tim too
    Malnutrition.

    I was initially inclined to dismiss that glib response, but you have made some interesting comments in your short time posting at unz.com so perhaps it is worth engaging.

    Do you think malnutrition is the only factor explaining the sea of red in APilgrim’s graphic? Surely there are other countries that are malnourished? Let’s take a look: https://ourworldindata.org/hunger-and-undernourishment

    (edit: looks like SVG files do not embed here, click through to see map)

    https://d33wubrfki0l68.cloudfront.net/fdb5a6c43da0d5bc865808e9767c97dddc0fa053/6ea73/exports/global-hunger-index-e3bcb6c247c02f1e57c637d2d9c7fe75_v1_850x600.svg

    Try another:

    http://cdn3.chartsbin.com/chartimages/l_3355_474b042fc1dc129b6e492440c7bd78aa

    The choropleths are somewhat similar, but there is a notable disconnect between the Subcontinent and sub-Saharan Africa. How do you explain that?

    Read More
    • Replies: @Tim too
    res, I have no idea, it was half way between a question, and other info I've seen about malnutrition demographics. No doubt malnutrition occurs else where. Lots of puzzling data. Nutrition is just one piece of the puzzle.
    , @Okechukwu

    Do you think malnutrition is the only factor explaining the sea of red in APilgrim’s graphic? Surely there are other countries that are malnourished? Let’s take a look: https://ourworldindata.org/hunger-and-undernourishment
     
    The sea of red is fraudulent, dummy. Can you identify which Africans were IQ tested? Who tested them? Sample size? Methodology? Where? When? How?

    Jesus, how dumb can you people be?
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  54. Tim too says:
    @res
    I was initially inclined to dismiss that glib response, but you have made some interesting comments in your short time posting at unz.com so perhaps it is worth engaging.

    Do you think malnutrition is the only factor explaining the sea of red in APilgrim's graphic? Surely there are other countries that are malnourished? Let's take a look: https://ourworldindata.org/hunger-and-undernourishment

    (edit: looks like SVG files do not embed here, click through to see map)

    https://d33wubrfki0l68.cloudfront.net/fdb5a6c43da0d5bc865808e9767c97dddc0fa053/6ea73/exports/global-hunger-index-e3bcb6c247c02f1e57c637d2d9c7fe75_v1_850x600.svg

    Try another:

    http://cdn3.chartsbin.com/chartimages/l_3355_474b042fc1dc129b6e492440c7bd78aa

    The choropleths are somewhat similar, but there is a notable disconnect between the Subcontinent and sub-Saharan Africa. How do you explain that?

    res, I have no idea, it was half way between a question, and other info I’ve seen about malnutrition demographics. No doubt malnutrition occurs else where. Lots of puzzling data. Nutrition is just one piece of the puzzle.

    Read More
    • Replies: @res
    Agreed nutrition is part of the puzzle. I do think the environment (e.g. nutrition, disease, lack of education) is a significant contributor to the extremely low average IQs seen in Africa.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  55. Okechukwu says:
    @APilgrim
    Lynn and Vanhanen: Maroon is below 65, Purple is over 105: Deal with it.
    . https://upload.wikimedia.org/wikipedia/commons/5/5e/National_IQ_per_country_-_estimates_by_Lynn_and_Vanhanen_2006.png

    Lynn and Vanhanen: Maroon is below 65, Purple is over 105: Deal with it.

    I’m familiar with the chart. What’s shocking is that someone’s dumb enough to believe it. Rather ironic in a discussion of intelligence.

    Funny how African athletes are able to compete in the normal Olympics rather than the Special Olympics.

    Mental retardation is cognitive limitation as characterized by scores greater than 2 standard deviations below the mean on a valid intelligence quotient (IQ) measure, with limitation of adaptive function in communication, self‐care, daily living skills at home or in the community, or social skills.

    https://www.sciencedirect.com/topics/neuroscience/mental-retardation

    Not only do Africans not have any of these issues, when they immigrate to the west they outperform whites.

    African immigrants are more educated than most — including people born in U.S.

    Batalova’s research found that of the 1.4 million who are 25 and older, 41% have a bachelor’s degree, compared with 30% of all immigrants and 32% of the U.S.-born population. Of the 19,000 U.S. immigrants from Norway — a country Trump reportedly told lawmakers is a good source of immigrants — 38% have college educations.

    The New American Economy study found that 1 in 3 of these undergraduate degrees were focused on science, technology, engineering and math — “training heavily in demand by today’s employers.”

    That report also found that African immigrants were significantly more likely to have graduate degrees. A total of 16% had a master’s degree, medical degree, law degree or a doctorate, compared with 11% of the U.S.-born population.

    http://www.latimes.com/world/africa/la-fg-global-african-immigrants-explainer-20180112-story.html

    I don’t know who’s more stupid, Lynn and Vanhanen or people like you who accept their “research.”

    Read More
    • Replies: @Anon
    > African immigrants are more educated than most — including people born in U.S.

    Lynn and Vanhanen showed the IQ in the native countries. You tried to counter that with specially selected elite African immigrants with that?? You do not seem to understand selection bias.

    You are a pathetic loser.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  56. Okechukwu says:
    @res
    I was initially inclined to dismiss that glib response, but you have made some interesting comments in your short time posting at unz.com so perhaps it is worth engaging.

    Do you think malnutrition is the only factor explaining the sea of red in APilgrim's graphic? Surely there are other countries that are malnourished? Let's take a look: https://ourworldindata.org/hunger-and-undernourishment

    (edit: looks like SVG files do not embed here, click through to see map)

    https://d33wubrfki0l68.cloudfront.net/fdb5a6c43da0d5bc865808e9767c97dddc0fa053/6ea73/exports/global-hunger-index-e3bcb6c247c02f1e57c637d2d9c7fe75_v1_850x600.svg

    Try another:

    http://cdn3.chartsbin.com/chartimages/l_3355_474b042fc1dc129b6e492440c7bd78aa

    The choropleths are somewhat similar, but there is a notable disconnect between the Subcontinent and sub-Saharan Africa. How do you explain that?

    Do you think malnutrition is the only factor explaining the sea of red in APilgrim’s graphic? Surely there are other countries that are malnourished? Let’s take a look: https://ourworldindata.org/hunger-and-undernourishment

    The sea of red is fraudulent, dummy. Can you identify which Africans were IQ tested? Who tested them? Sample size? Methodology? Where? When? How?

    Jesus, how dumb can you people be?

    Read More
    • LOL: res
    • Replies: @res

    Can you identify which Africans were IQ tested? Who tested them? Sample size? Methodology? Where? When? How?
     
    The most comprehensive resource for worldwide country IQ information which I know of is David Becker's NIQ-dataset available at https://www.researchgate.net/project/Worlds-IQ
    In version 1.3 he gives the details of 670 studies worldwide, including 10 in Nigeria alone. If you are truly interested in the answers to those questions, then download the database and follow the references given in the spreadsheet entries for each study.

    If you were actually paying attention, this dataset has been discussed multiple times on Dr. Thompson's blog:
    http://www.unz.com/jthompson/richard-lynn-intelligence-database/
    http://www.unz.com/jthompson/the-worlds-iq-86/
    http://www.unz.com/jthompson/world-iq-latest-update/

    P.S. For those who are interested in this dataset, I should note that a corrected version of version 1.3 was uploaded just yesterday. The dataset itself was not available as of the most recent discussion here in May (final link above, see my comment and David Becker's response). If David Becker happens to read this, thank you! (you were right on the schedule you gave two months ago)
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  57. hyperbola says:
    @j2
    hyperbola writes:
    "(why should “additive” be linear – complex interdependence is physiologically much more reasonable)"

    This is a mathematical issue. Most of these complex interdependences are continuous, differentiable functions. A small effect is then f(p)+f '(p)dx, it is linear (i.e., effects are additive). This is true for small effects and it explains some part why such additive constructs like a polygenic score cannot explain much of the effect: it is additive (=linear) only for small changes. In order to find out the whole function (all partial derivatives) would require very much study.

    The mathematically more reasonable way to evaluate the effect of SNPs to IQ would be to show that an SNP has a small effect rather than no effect (as we cannot find out the whole effect without knowing the function) and then scan over all genes. The set of genes that have any effect account for the whole genetic effect, estimated to be 60-80%. Then you can assume that the effects are distributed in some way, like normally (a few have larger effects, most intermediate) and to calculate the average effect.

    The assumption of linearity over small changes (dx) is not warranted by any real evidence – a step function rather than a continuous function might be more characteristic of physiological functions. Imagine 2 proteins that interact directly. Their interaction might be turned off completely by a change that affects either one of them. Whether this interaction is dominant for “our” target “observable” remains an open question.

    Maybe we have to think about this in completely different ways. The reason that we cannot identify the dominant “mutations” for “IQ” is that they are lethal and we never have them included in the samples. The remaining “mutations” observed in our samples have only rather peripheral effects (to avoid lethality), may well be part of highly parallel partly redundant “functions” and will be essentially useless for prediction of things like “IQ”.

    Speculation, but maybe it serves to imagine that the “canned” analyses currently being applied to these kinds of problems is far from being definitive.

    Read More
    • Replies: @j2
    We can think like this. There are different alleles (let them be SNPs for simplicity). Some lucky combination causes high genetic IQ and unlucky low. But in a population study when measuring IQ of all people who have a certain allele, we see only a small rising or sinking effect. If this population has more of the allele, we see a stronger effect. Here is a continuous and differentiable function of the average IQ of the population with respect to the allele frequency. For small changes of allele frequency the effect is nearly linear. Summing over all IQ influencing alleles, we have a polygenic score, which then is a linear sum of the allele frequency variables. Comparing one population to another fairly similar population we can calculate the difference from the first order linear approximation. The function is not of an individual when there is one allele or another and it is discrete. It is of allele frequency in a large population, a continuous variable.

    Then think about how much of the IQ variance we should expect that this additive polygenic score should predict? Only a percent or so, because it is a first order approximation. To give you an example. Assume we have gene A with alleles A1, A2 and gene B with alleles B1 and B2. Assume A1B1 gives a very high IQ, say 180 and any other combination gives 100. In a population with 1% of A1 and 1% of B1 we get 1/10000 A1b1 and the average IQ effect of A1 is 80/10000. B1 has the same effect 80/10000. The additive effect is 160/10000, but in this fictive population with only two IQ genes and 4 alleles, 100% of IQ differences is determined by these alleles, and the IQ differences are that the IQ of two individuals can differ by 80. So, how much polygenic score explains of IQ differences does not indicate how much of IQ differences the SNPs in the polygenic score explain.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  58. hyperbola says:
    @utu

    This is why I prefer to speak about genes rather than SNPs.
     
    If this is true "that over 98% of the human genome does not encode protein sequences" (wiki) then out of 10 million SNPs only 200,000 belong to the coding regions. In this paper (See my comment #16) to obtain 11-13% variance explanation they effectively used 1 million of SNPs while with 1,271 "lead" SNPs they explained only 3.9%. This means that 800,000 SNPs they used in their polygenic score were form the noncoding regions. Not surprisingly commenter res wrote in his comment #6:

    Supplementary Table 10 has a list of the causal SNPs. One thing that surprises me is how few of them are coding SNPs.
     
    (BTW, where the Supplementary material is available?)

    It is possibly that the dichotomy of coding/non-coding DNA is false to some extent and there is much more going on than what was assumed so far in genetics. But if it was true that only SNP's from the coding regions should suffice to explain any trait, then if all 200,000 SNPs from coding regions were already included in the polygenic score and they still could not replicate the twin based heritability of some trait, then the case should be closed with a conclusion that the objections raised by the skeptics of twin based heritability must have been correct that twin based heritability overestimates actual heritability.

    Though, as I wrote in one of the earlier comments, the polygenic sore does not have to be linear and additive. Nonlinearity, afaik, was not tested though Hsu wrote one paper last year where he proposed to use 2nd order parabolic nonlinearity but, afaik, he did not implement it. I think the reason he wanted to do it was because his height linear predictor function could not reach twin based heritability even though he used 10,000 SNPs.

    I do not understand why people who make the usual objections to DNA studies of intelligence do not bring the issue of 1,000,000 SNPs used in this paper. For God's sake there is not much more left in coding or nearby regions and all they could get was 11-13%. One reason is that the issue of 1,000,000 SNPs is well hidden in FAQ part and does not appear in the main body of the paper. The abstract of the paper should have been rewritten: 1,271 lead SNPs explain 3.9% of variance. The explained variance increases to 11-13% by adding extra 1,000,000 SNPs into the polygenic score. to avoid gross misrepresentation of the results.

    (1) There may still be further effects associated with non-coding regions. Chromatin structure and its packaging in the nucleus is one obvious area for such effects. There are some studies that attempt to look at changes in the spatial organization of nuclear chromatin as a function of environmental conditions for a cell. It includes things like changes in the interaction of chromatin with the nuclear envelope (hence potential changes in expression of different regions of chromatin).

    (2) Many SNPs may be functionally neutral, i.e. any effect they produce is to small to be observed.

    These studies and their “canned” forms of analysis are wearing out my patience to keep reading them!

    Read More
    • Replies: @utu

    These studies and their “canned” forms of analysis are wearing out my patience to keep reading them!
     
    You should look at them and try to unpack them. There are legitimate questions. Since twin studies show that IQ score is highly heritable then a predictor function of IQ score based on SNPs should be possible to construct. It is possible as some critics say that the twin studies base heritability is overestimated but I think nobody believes that it should be zero. Therefore a nonzero function mapping genes onto the scale of IQ scores must exist. Because mathematically the problem is insanely undetermined (7 billion people and almost infinite number of combinations out of 10 million SNP set) the problem is not finding such a function but finding the one that is not spurious as the result of overfitting.

    I have read somewhere that 88% GWAS hits usually come from the noncoding part of DNA in studies of all kinds of different traits. So what is the biological mechanism how noncoding SPNs affect a trait? This is an importnt problem which however should not stop us from looking for mathematically constructed functions even if it includes SNPs that biologists do not know what is their role.

    IQ score, intelligence, cognitive abilities , educational attainments pose additional problem as they all are related to some not well defined trait which certainly is very complex. It is not like height that has a good definition and can be measured with a physical scale. Nevertheless twin studies suggest that heritability within our type of society is around 50-70%. Obviously if we could include twins raised by wolves the result would be much different.

    As far as reading the papers it is really hard because there is way too much of unnecessary professional jargon that obfuscates even if not intentionally, though I have some doubts about how unintentional it is. Look at this paper and you will see how scrupulous they are about reporting P-values of anything and everything. This is mostly BS. But on the other hand when it comes to spell out the polygenic score definition that contains 1,000,000 SPNs they are silent. They have mentioned it en passant in the FAQ section only. Or look at a brief paragraph when they used the predictor (white) function on Afro-American population. They do not explain which predictor function they used. Was it the one that produced 3.9%. R^2 or the one that produced 11-13% R^2 on white population? They also could have told us what was the offset bias between the two populations.

    I would not mind if there was a higher authority that could exercise its power of subpoena over the cliques of scientist and have them undergo interrogations where they would have to give some explanations. I could volunteer for the position of Torquemada.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  59. Okechukwu says:
    @res

    No one denies individual differences in intelligence. This is strawmanning at its best (or worst).
     
    I specifically referred to denying the influence of genetics on intelligence. In other words, you strawmanned me by leaving out the "genetics" qualifier. But you are right that your first sentence was an example of strawmanning at its worst (that was what you meant, right?).

    As for no one denying the influence of genetics of on intelligence, if that is so then why did someone feel the need to write this article in 2017? Yes, There Is a Genetic Component to Intelligence https://www.thecut.com/2017/05/genetics-intelligence.html


    Given the plasticity of IQ and given the tremendous within race variation and between race overlap, you’d have to be pretty illiterate scientifically to claim that genes play a role in the black/white IQ gap. Moreover, a genetic interpretation wouldn’t necessarily support the conclusion you are pining for. If you followed the debate that was instigated by the Vox article, you’ll recall that Ezra Klein called James Flynn a few days before his encounter with Sam Harris. Flynn went on to say that it is entirely possible that the 10-point IQ gap reflects a 12 point environmental difference and a negative 2 point genetic difference. That is, blacks could very well have an inherent advantage in intelligence that is overwhelmed by environmental factors. This to me is more plausible than the alternative.
     
    That was some Turkheimer worthy obfuscation. Let's make this simple. Yes or no, do you believe there is exactly zero genetic contribution to the Black/White IQ difference in the US? Any answer other than an explicit "yes" indicates agreement with the conclusion I quoted from The Bell Curve. To repeat the most important part: "It seems highly likely to us that both genes and the environment have something to do with racial differences."

    On what basis do you think a black genetic advantage in intelligence is more likely than the opposite? I believe the opposite is more likely (has higher probability) because the required environmental difference is smaller if the phenotypic difference has the same sign as the genetic difference. A smaller required environmental difference is more likely.


    There’s nothing you can present that can’t be picked apart and comprehensively rebutted.
     
    The SNP frequencies differ between the races for the first (and likely almost all) IQ SNPs found. That is reality. Sorry you don't like it.

    You’re holding a house of cards, bruh. Even a baby’s breath can blown it down.

     

    Now that was some world class argumentation. I tremble before the awesomeness of your debating skills. /sarc

    To a rational, unbiased observer the second article served only to further bury Murray and his racist pseudoscience with much more detail than the first.
     
    Because the failure of the three authors to agree on a single response demonstrates just how compelling their arguments are. LOL!

    I specifically referred to denying the influence of genetics on intelligence.

    No one has ever denied genetic or inherited intelligence. Even prehistoric humans knew that attributes like intelligence can be transmitted from parent to offspring.

    Yes or no, do you believe there is exactly zero genetic contribution to the Black/White IQ difference in the US?

    It’s pointless. Nigerian-Americans are reputed to have an IQ of 110. Is their nearly 1 standard deviation advantage over whites effectuated genetically? There are all kinds of group IQ disparities. Why are you insistent on a genetic cause only with respect to the black/white gap in the US? Is the southern and northern white American IQ gap genetically based? How about the Catholic and Protestant Irish IQ gap? Or the Japanese and ethnic Korean IQ gap in Japan? Etc., Etc., Etc.

    On what basis do you think a black genetic advantage in intelligence is more likely than the opposite?

    Because after 400 years of oppression and cultural and social ostracization they are only 9 or 10 points behind. Flynn is too conservative. Whites should be 20-30 points ahead rather than a mere 9.

    The SNP frequencies differ between the races for the first (and likely almost all) IQ SNPs found. That is reality. Sorry you don’t like it.

    If you weren’t an idiot you would be highly skeptical of that data, given that smart and genius black people do exist … Duh. Btw, there is no such thing as an “IQ SNP.”

    Because the failure of the three authors to agree on a single response demonstrates just how compelling their arguments are. LOL!

    That’s only your opinion. Those three authors are firmly anchored within the mainstream of scientific thought. You, Sailer and Murray are the outliers. Furthermore, we know that your thoughts and reactions are motivated by varying strains of white nationalism/supremacism. So you are not credible actors. Therefore your conclusions are unreliable.

    Read More
    • Replies: @res

    No one has ever denied genetic or inherited intelligence.
     
    I truly enjoy it when you give a partial quote of my comment then make an assertion which I specifically responded to immediately following the text you quoted. It is almost like I am thinking a move ahead of you.

    Is their nearly 1 standard deviation advantage over whites effectuated genetically?
     
    If that difference truly exists as you say (data? I love your selective demands for rigor), then quite possibly. It would make an interesting study (maybe you could even convince someone to do it given the PC conclusion). Given how far they are outliers from Nigeria itself I would suspect regression to the mean indicating a significant positive environmental component though.

    And since Nigerian-Americans are so superior, I think they should volunteer to relinquish all of their affirmative action slots in favor of slave descended African-Americans.

    Why are you insistent on a genetic cause only with respect to the black/white gap in the US?
     
    I'm not. Please don't confuse the caricature of me you have in your mind (or the strawmen you attempt to construct) with my actual beliefs.

    If I seem to place too much emphasis on the US black/white gap that is because of the poisonous influence "disparate impact" has on the thinking here. Assuming that any difference in group representation or outcomes is due to racism is ridiculous.

    Is the southern and northern white American IQ gap genetically based? How about the Catholic and Protestant Irish IQ gap? Or the Japanese and ethnic Korean IQ gap in Japan?
     
    All good questions. Hard to say without more data than we have. Part of the problem is those gaps are all smaller than the US black/white gap so harder to observe without better data than is generally available.

    Because after 400 years of oppression and cultural and social ostracization they are only 9 or 10 points behind. Flynn is too conservative. Whites should be 20-30 points ahead rather than a mere 9.
     
    On what basis do you make those particular numerical estimates? Is there some kind of equation like
    1 microagression = 0.001 IQ points

    If you weren’t an idiot you would be highly skeptical of that data, given that smart and genius black people do exist …
     
    Learn some statistics.

    Duh
     
    Are you 12 years old?

    Those three authors are firmly anchored within the mainstream of scientific thought. You, Sailer and Murray are the outliers.
     
    That is certainly true. I guess we will need to come back to this conversation after the evidence continues to roll in for a few more years. It is charming the way you think consensus determines truth. You might want to spend some time studying the history of science.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  60. APilgrim says:
    @Tim too
    Malnutrition.

    Most of the Purple Areas experienced STARVATION, during most of the 20th Century.

    During WWII, the entire Purple area was malnourished.

    North Koreans are still grossly malnourished. NORKS have TWICE the IQ of Blacks.

    Read More
    • Replies: @Tim too
    well, the graphic in the post has no date on it, so how is anyone supposed to know? does it apply now? or during the mass starvations? WWII was a long time ago. Malnutrition in infancy/childhood does have lasting effects on intelligence. And who's data are you going by for the DPRK malnutrition? And who's IQ test is it? IQ depends on the test writer/creator as well as the test subjects. Remember, most people at select times past could not read, demographic IQ is not consistent over time. So the IQ for Russia eg, would be very different in say 1800, 1900, and 2000. That map of yours over time would show lots more red in many more places in the past.

    Over-interpretation of data is a problem. More refinement, definition/resolution of data is needed.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  61. utu says:
    @j2
    hyperbola writes:
    "(why should “additive” be linear – complex interdependence is physiologically much more reasonable)"

    This is a mathematical issue. Most of these complex interdependences are continuous, differentiable functions. A small effect is then f(p)+f '(p)dx, it is linear (i.e., effects are additive). This is true for small effects and it explains some part why such additive constructs like a polygenic score cannot explain much of the effect: it is additive (=linear) only for small changes. In order to find out the whole function (all partial derivatives) would require very much study.

    The mathematically more reasonable way to evaluate the effect of SNPs to IQ would be to show that an SNP has a small effect rather than no effect (as we cannot find out the whole effect without knowing the function) and then scan over all genes. The set of genes that have any effect account for the whole genetic effect, estimated to be 60-80%. Then you can assume that the effects are distributed in some way, like normally (a few have larger effects, most intermediate) and to calculate the average effect.

    Keep in mind that it is not even possible to writes something like this “f(p)+f ‘(p)dx” in the case of the polygenic score. The independent variables are binary. One either have a given allele of SNP or not. The linearity in question concerns multiple variables not a single variable. The nonlinearity in terms of one variable does not even makes sense because variable is basically binary. If x and y are two different SNPs GWAS can approximate the answer to what is the difference in trait when x=0 or 1, i.e., when you have a given allele of the SNP or not. The same can be done for SNP y. But it is harder to answer whether the effect of x and y together is a sum of effects due to separate effects of x and of y. And when you have more and more SNPs that you want to include in the polygenic score it becomes harder and harder to disentangle their non liner interactions. In case of two binary variables x and y you can write polygenic score as: ax+by or ax+by+cxy. In the second case the nonlinear term xy, which kicks in only when both x and y are non zero, takes care of the possibility that the effect of two SNPs is not equal to the sum of the effects. This formulation makes it possible to account for the case when x and y individually have no effect (a=b=0) but in tandem they do. More or less this is how it should be done. The problem is that when you consider n SNPs in a linear case you have n variables but in nonlinear case with mixed terms you have additionally n*(n-1)/2 terms to consider. In the case of 1,000 SNPs you’ll end up with 500,000 variables to be determined.

    Read More
    • Replies: @Chainsaw1
    > Keep in mind that it is not even possible to writes something like this “f(p)+f ‘(p)dx” in the case of the polygenic score. The independent variables are binary.

    Another stupid ignorant fool. Read the freaking supp before opening your big mouth. In PGS the independent variables are computed from the fraction distribution of the SNP variants, it can be as fine as you want with increasing sample size.


    1.5. Association Analyses

    Cohorts were asked to estimate this regression equation for each measured SNP:

    EduYears = Bo + B1*SNP + PC gamma + B*alpha + X + e, (1.1)

    where SNP is the allele dose of the SNP;

     

    https://www.biostars.org/p/75689/

    Assuming that you have a SNP: A/B and your genotype probablities are:

    A/A : 0.1
    A/B: 0.4
    B/B: 0.5

    (They should all sum to 1.0)

    Then the dosage for this SNP is: 0*A/A + 1*A/B + 2*B/B = 0.4 + 2*0.5 = 1.4

    So the maximum dosage you can get is 2.0 (that is if the genotype probabilies of 0 for A/A, A/B and 1.0 for B/B)

     

    With bigger sample size for determining the dosage, the finer the dosage value. Now shut up your stupid big mouth.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  62. hyperbola says:
    @James Thompson
    Thanks for directing me to Turkheimer's remarks, which end thus:

    it does not show that differences in outcomes among racial and ethnic groups are genetically determined. In fact none of these things are any more likely today than they were yesterday.

    I think that this is wrong, in that every paper which shows a detailed link between the genetic code of one genetic group and an important outcome, like years of completed education, strengthens the likelihood that another genetic group might have a different code which leads to different outcomes. However, if it was impossible to show that there was any link in any genetic group with educational outcomes, then the chance of it accounting for racial differences in educational attainment would be much weaker. Currently, searching for genetic differences as an explanation for genetic group outcomes has become more tenable, not less.

    Unfortunately the pseudo-science in this paper does NOT support your conclusion.

    Currently, searching for genetic differences as an explanation for genetic group outcomes has become more tenable, not less.

    It seems to be time to exclude the “psychologists” from being examples of science.

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  63. res says:
    @Tim too
    res, I have no idea, it was half way between a question, and other info I've seen about malnutrition demographics. No doubt malnutrition occurs else where. Lots of puzzling data. Nutrition is just one piece of the puzzle.

    Agreed nutrition is part of the puzzle. I do think the environment (e.g. nutrition, disease, lack of education) is a significant contributor to the extremely low average IQs seen in Africa.

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  64. res says:
    @Okechukwu

    Do you think malnutrition is the only factor explaining the sea of red in APilgrim’s graphic? Surely there are other countries that are malnourished? Let’s take a look: https://ourworldindata.org/hunger-and-undernourishment
     
    The sea of red is fraudulent, dummy. Can you identify which Africans were IQ tested? Who tested them? Sample size? Methodology? Where? When? How?

    Jesus, how dumb can you people be?

    Can you identify which Africans were IQ tested? Who tested them? Sample size? Methodology? Where? When? How?

    The most comprehensive resource for worldwide country IQ information which I know of is David Becker’s NIQ-dataset available at https://www.researchgate.net/project/Worlds-IQ
    In version 1.3 he gives the details of 670 studies worldwide, including 10 in Nigeria alone. If you are truly interested in the answers to those questions, then download the database and follow the references given in the spreadsheet entries for each study.

    If you were actually paying attention, this dataset has been discussed multiple times on Dr. Thompson’s blog:

    http://www.unz.com/jthompson/richard-lynn-intelligence-database/

    http://www.unz.com/jthompson/the-worlds-iq-86/

    http://www.unz.com/jthompson/world-iq-latest-update/

    P.S. For those who are interested in this dataset, I should note that a corrected version of version 1.3 was uploaded just yesterday. The dataset itself was not available as of the most recent discussion here in May (final link above, see my comment and David Becker’s response). If David Becker happens to read this, thank you! (you were right on the schedule you gave two months ago)

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  65. res says:
    @Okechukwu

    I specifically referred to denying the influence of genetics on intelligence.
     
    No one has ever denied genetic or inherited intelligence. Even prehistoric humans knew that attributes like intelligence can be transmitted from parent to offspring.

    Yes or no, do you believe there is exactly zero genetic contribution to the Black/White IQ difference in the US?
     
    It's pointless. Nigerian-Americans are reputed to have an IQ of 110. Is their nearly 1 standard deviation advantage over whites effectuated genetically? There are all kinds of group IQ disparities. Why are you insistent on a genetic cause only with respect to the black/white gap in the US? Is the southern and northern white American IQ gap genetically based? How about the Catholic and Protestant Irish IQ gap? Or the Japanese and ethnic Korean IQ gap in Japan? Etc., Etc., Etc.

    On what basis do you think a black genetic advantage in intelligence is more likely than the opposite?
     
    Because after 400 years of oppression and cultural and social ostracization they are only 9 or 10 points behind. Flynn is too conservative. Whites should be 20-30 points ahead rather than a mere 9.

    The SNP frequencies differ between the races for the first (and likely almost all) IQ SNPs found. That is reality. Sorry you don’t like it.
     
    If you weren't an idiot you would be highly skeptical of that data, given that smart and genius black people do exist ... Duh. Btw, there is no such thing as an "IQ SNP."

    Because the failure of the three authors to agree on a single response demonstrates just how compelling their arguments are. LOL!
     
    That's only your opinion. Those three authors are firmly anchored within the mainstream of scientific thought. You, Sailer and Murray are the outliers. Furthermore, we know that your thoughts and reactions are motivated by varying strains of white nationalism/supremacism. So you are not credible actors. Therefore your conclusions are unreliable.

    No one has ever denied genetic or inherited intelligence.

    I truly enjoy it when you give a partial quote of my comment then make an assertion which I specifically responded to immediately following the text you quoted. It is almost like I am thinking a move ahead of you.

    Is their nearly 1 standard deviation advantage over whites effectuated genetically?

    If that difference truly exists as you say (data? I love your selective demands for rigor), then quite possibly. It would make an interesting study (maybe you could even convince someone to do it given the PC conclusion). Given how far they are outliers from Nigeria itself I would suspect regression to the mean indicating a significant positive environmental component though.

    And since Nigerian-Americans are so superior, I think they should volunteer to relinquish all of their affirmative action slots in favor of slave descended African-Americans.

    Why are you insistent on a genetic cause only with respect to the black/white gap in the US?

    I’m not. Please don’t confuse the caricature of me you have in your mind (or the strawmen you attempt to construct) with my actual beliefs.

    If I seem to place too much emphasis on the US black/white gap that is because of the poisonous influence “disparate impact” has on the thinking here. Assuming that any difference in group representation or outcomes is due to racism is ridiculous.

    Is the southern and northern white American IQ gap genetically based? How about the Catholic and Protestant Irish IQ gap? Or the Japanese and ethnic Korean IQ gap in Japan?

    All good questions. Hard to say without more data than we have. Part of the problem is those gaps are all smaller than the US black/white gap so harder to observe without better data than is generally available.

    Because after 400 years of oppression and cultural and social ostracization they are only 9 or 10 points behind. Flynn is too conservative. Whites should be 20-30 points ahead rather than a mere 9.

    On what basis do you make those particular numerical estimates? Is there some kind of equation like
    1 microagression = 0.001 IQ points

    If you weren’t an idiot you would be highly skeptical of that data, given that smart and genius black people do exist …

    Learn some statistics.

    Duh

    Are you 12 years old?

    Those three authors are firmly anchored within the mainstream of scientific thought. You, Sailer and Murray are the outliers.

    That is certainly true. I guess we will need to come back to this conversation after the evidence continues to roll in for a few more years. It is charming the way you think consensus determines truth. You might want to spend some time studying the history of science.

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  66. j2 says:
    @hyperbola
    The assumption of linearity over small changes (dx) is not warranted by any real evidence - a step function rather than a continuous function might be more characteristic of physiological functions. Imagine 2 proteins that interact directly. Their interaction might be turned off completely by a change that affects either one of them. Whether this interaction is dominant for "our" target "observable" remains an open question.

    Maybe we have to think about this in completely different ways. The reason that we cannot identify the dominant "mutations" for "IQ" is that they are lethal and we never have them included in the samples. The remaining "mutations" observed in our samples have only rather peripheral effects (to avoid lethality), may well be part of highly parallel partly redundant "functions" and will be essentially useless for prediction of things like "IQ".

    Speculation, but maybe it serves to imagine that the "canned" analyses currently being applied to these kinds of problems is far from being definitive.

    We can think like this. There are different alleles (let them be SNPs for simplicity). Some lucky combination causes high genetic IQ and unlucky low. But in a population study when measuring IQ of all people who have a certain allele, we see only a small rising or sinking effect. If this population has more of the allele, we see a stronger effect. Here is a continuous and differentiable function of the average IQ of the population with respect to the allele frequency. For small changes of allele frequency the effect is nearly linear. Summing over all IQ influencing alleles, we have a polygenic score, which then is a linear sum of the allele frequency variables. Comparing one population to another fairly similar population we can calculate the difference from the first order linear approximation. The function is not of an individual when there is one allele or another and it is discrete. It is of allele frequency in a large population, a continuous variable.

    Then think about how much of the IQ variance we should expect that this additive polygenic score should predict? Only a percent or so, because it is a first order approximation. To give you an example. Assume we have gene A with alleles A1, A2 and gene B with alleles B1 and B2. Assume A1B1 gives a very high IQ, say 180 and any other combination gives 100. In a population with 1% of A1 and 1% of B1 we get 1/10000 A1b1 and the average IQ effect of A1 is 80/10000. B1 has the same effect 80/10000. The additive effect is 160/10000, but in this fictive population with only two IQ genes and 4 alleles, 100% of IQ differences is determined by these alleles, and the IQ differences are that the IQ of two individuals can differ by 80. So, how much polygenic score explains of IQ differences does not indicate how much of IQ differences the SNPs in the polygenic score explain.

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  67. utu says:
    @hyperbola
    (1) There may still be further effects associated with non-coding regions. Chromatin structure and its packaging in the nucleus is one obvious area for such effects. There are some studies that attempt to look at changes in the spatial organization of nuclear chromatin as a function of environmental conditions for a cell. It includes things like changes in the interaction of chromatin with the nuclear envelope (hence potential changes in expression of different regions of chromatin).

    (2) Many SNPs may be functionally neutral, i.e. any effect they produce is to small to be observed.


    These studies and their "canned" forms of analysis are wearing out my patience to keep reading them!

    These studies and their “canned” forms of analysis are wearing out my patience to keep reading them!

    You should look at them and try to unpack them. There are legitimate questions. Since twin studies show that IQ score is highly heritable then a predictor function of IQ score based on SNPs should be possible to construct. It is possible as some critics say that the twin studies base heritability is overestimated but I think nobody believes that it should be zero. Therefore a nonzero function mapping genes onto the scale of IQ scores must exist. Because mathematically the problem is insanely undetermined (7 billion people and almost infinite number of combinations out of 10 million SNP set) the problem is not finding such a function but finding the one that is not spurious as the result of overfitting.

    I have read somewhere that 88% GWAS hits usually come from the noncoding part of DNA in studies of all kinds of different traits. So what is the biological mechanism how noncoding SPNs affect a trait? This is an importnt problem which however should not stop us from looking for mathematically constructed functions even if it includes SNPs that biologists do not know what is their role.

    IQ score, intelligence, cognitive abilities , educational attainments pose additional problem as they all are related to some not well defined trait which certainly is very complex. It is not like height that has a good definition and can be measured with a physical scale. Nevertheless twin studies suggest that heritability within our type of society is around 50-70%. Obviously if we could include twins raised by wolves the result would be much different.

    As far as reading the papers it is really hard because there is way too much of unnecessary professional jargon that obfuscates even if not intentionally, though I have some doubts about how unintentional it is. Look at this paper and you will see how scrupulous they are about reporting P-values of anything and everything. This is mostly BS. But on the other hand when it comes to spell out the polygenic score definition that contains 1,000,000 SPNs they are silent. They have mentioned it en passant in the FAQ section only. Or look at a brief paragraph when they used the predictor (white) function on Afro-American population. They do not explain which predictor function they used. Was it the one that produced 3.9%. R^2 or the one that produced 11-13% R^2 on white population? They also could have told us what was the offset bias between the two populations.

    I would not mind if there was a higher authority that could exercise its power of subpoena over the cliques of scientist and have them undergo interrogations where they would have to give some explanations. I could volunteer for the position of Torquemada.

    Read More
    • Replies: @j2
    "So what is the biological mechanism how noncoding SPNs affect a trait?"

    I guess it usually is a noncoding part. Coding parts code different proteins, which may change something but normally to the worse. Non-coding part contains control parts that modify how long some protein is produced. So, for instance, the brain gets bigger if the growing stage (when some proteins are produced) is longer. Non-coding is not the same as junk DNA.
    , @res

    I have read somewhere that 88% GWAS hits usually come from the noncoding part of DNA in studies of all kinds of different traits. So what is the biological mechanism how noncoding SPNs affect a trait?
     
    You might want to check out this paper: Mining the Unknown: Assigning Function to Noncoding Single Nucleotide Polymorphisms
    https://www.cell.com/trends/genetics/pdf/S0168-9525(16)30147-0.pdf (paywalled)
    DOI 10.1016/j.tig.2016.10.008

    This paper might be even better: Beyond GWASs: Illuminating the Dark Road from Association to Function
    https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3824120/

    Abstract (emphasis mine):

    Genome-wide association studies (GWASs) have enabled the discovery of common genetic variation contributing to normal and pathological traits and clinical drug responses, but recognizing the precise targets of these associations is now the major challenge. Here, we review recent approaches to the functional follow-up of GWAS loci, including fine mapping of GWAS signal(s), prioritization of putative functional SNPs by the integration of genetic epidemiological and bioinformatic methods, and in vitro and in vivo experimental verification of predicted molecular mechanisms for identifying the targeted genes. The majority of GWAS-identified variants fall in noncoding regions of the genome. Therefore, this review focuses on strategies for assessing likely mechanisms affected by noncoding variants; such mechanisms include transcriptional regulation, noncoding RNA function, and epigenetic regulation. These approaches have already accelerated progress from genetic studies to biological knowledge and might ultimately guide the development of prognostic, preventive, and therapeutic measures.
     
    , @hyperbola

    I have read somewhere that 88% GWAS hits usually come from the noncoding part of DNA in studies of all kinds of different traits. So what is the biological mechanism how noncoding SPNs affect a trait?
     
    For humans the estimates seem to be that protein-coding regions encompass on the order of 1% of DNA. There are other regions that are used to produce various important RNAs such as tRNA, microRNA, etc. Still other regions seem to be at least transcribed although no known (direct) functions are known and they may not be conserved evolutionarily. Other regions may show epigenetic effects (DNA methylation), but without identification of a specific functional activity. This is a very complex subject - you might find this useful.

    My guess would be that many of the "junk" regions have to do with the spatial structure and packaging of DNA and hence on expression of different protein-coding regions.

    Defining functional DNA elements in the human genome
    https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4035993/

    As for 10 million SNPs, this study only measured 1 million people. Can we expect 7000 times more SNPs when we expand to 7 billion humans? We already know that many of the SNPs seem to be in DNA regions that are not under evolutionary selection (see paper).
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  68. j2 says:
    @utu

    These studies and their “canned” forms of analysis are wearing out my patience to keep reading them!
     
    You should look at them and try to unpack them. There are legitimate questions. Since twin studies show that IQ score is highly heritable then a predictor function of IQ score based on SNPs should be possible to construct. It is possible as some critics say that the twin studies base heritability is overestimated but I think nobody believes that it should be zero. Therefore a nonzero function mapping genes onto the scale of IQ scores must exist. Because mathematically the problem is insanely undetermined (7 billion people and almost infinite number of combinations out of 10 million SNP set) the problem is not finding such a function but finding the one that is not spurious as the result of overfitting.

    I have read somewhere that 88% GWAS hits usually come from the noncoding part of DNA in studies of all kinds of different traits. So what is the biological mechanism how noncoding SPNs affect a trait? This is an importnt problem which however should not stop us from looking for mathematically constructed functions even if it includes SNPs that biologists do not know what is their role.

    IQ score, intelligence, cognitive abilities , educational attainments pose additional problem as they all are related to some not well defined trait which certainly is very complex. It is not like height that has a good definition and can be measured with a physical scale. Nevertheless twin studies suggest that heritability within our type of society is around 50-70%. Obviously if we could include twins raised by wolves the result would be much different.

    As far as reading the papers it is really hard because there is way too much of unnecessary professional jargon that obfuscates even if not intentionally, though I have some doubts about how unintentional it is. Look at this paper and you will see how scrupulous they are about reporting P-values of anything and everything. This is mostly BS. But on the other hand when it comes to spell out the polygenic score definition that contains 1,000,000 SPNs they are silent. They have mentioned it en passant in the FAQ section only. Or look at a brief paragraph when they used the predictor (white) function on Afro-American population. They do not explain which predictor function they used. Was it the one that produced 3.9%. R^2 or the one that produced 11-13% R^2 on white population? They also could have told us what was the offset bias between the two populations.

    I would not mind if there was a higher authority that could exercise its power of subpoena over the cliques of scientist and have them undergo interrogations where they would have to give some explanations. I could volunteer for the position of Torquemada.

    “So what is the biological mechanism how noncoding SPNs affect a trait?”

    I guess it usually is a noncoding part. Coding parts code different proteins, which may change something but normally to the worse. Non-coding part contains control parts that modify how long some protein is produced. So, for instance, the brain gets bigger if the growing stage (when some proteins are produced) is longer. Non-coding is not the same as junk DNA.

    Read More
    • Replies: @hyperbola

    “So what is the biological mechanism how noncoding SPNs affect a trait?”

    I guess it usually is a noncoding part.....
     
    A very complex subject.

    Defining functional DNA elements in the human genome
    https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4035993/
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  69. res says:
    @utu

    These studies and their “canned” forms of analysis are wearing out my patience to keep reading them!
     
    You should look at them and try to unpack them. There are legitimate questions. Since twin studies show that IQ score is highly heritable then a predictor function of IQ score based on SNPs should be possible to construct. It is possible as some critics say that the twin studies base heritability is overestimated but I think nobody believes that it should be zero. Therefore a nonzero function mapping genes onto the scale of IQ scores must exist. Because mathematically the problem is insanely undetermined (7 billion people and almost infinite number of combinations out of 10 million SNP set) the problem is not finding such a function but finding the one that is not spurious as the result of overfitting.

    I have read somewhere that 88% GWAS hits usually come from the noncoding part of DNA in studies of all kinds of different traits. So what is the biological mechanism how noncoding SPNs affect a trait? This is an importnt problem which however should not stop us from looking for mathematically constructed functions even if it includes SNPs that biologists do not know what is their role.

    IQ score, intelligence, cognitive abilities , educational attainments pose additional problem as they all are related to some not well defined trait which certainly is very complex. It is not like height that has a good definition and can be measured with a physical scale. Nevertheless twin studies suggest that heritability within our type of society is around 50-70%. Obviously if we could include twins raised by wolves the result would be much different.

    As far as reading the papers it is really hard because there is way too much of unnecessary professional jargon that obfuscates even if not intentionally, though I have some doubts about how unintentional it is. Look at this paper and you will see how scrupulous they are about reporting P-values of anything and everything. This is mostly BS. But on the other hand when it comes to spell out the polygenic score definition that contains 1,000,000 SPNs they are silent. They have mentioned it en passant in the FAQ section only. Or look at a brief paragraph when they used the predictor (white) function on Afro-American population. They do not explain which predictor function they used. Was it the one that produced 3.9%. R^2 or the one that produced 11-13% R^2 on white population? They also could have told us what was the offset bias between the two populations.

    I would not mind if there was a higher authority that could exercise its power of subpoena over the cliques of scientist and have them undergo interrogations where they would have to give some explanations. I could volunteer for the position of Torquemada.

    I have read somewhere that 88% GWAS hits usually come from the noncoding part of DNA in studies of all kinds of different traits. So what is the biological mechanism how noncoding SPNs affect a trait?

    You might want to check out this paper: Mining the Unknown: Assigning Function to Noncoding Single Nucleotide Polymorphisms
    https://www.cell.com/trends/genetics/pdf/S0168-9525(16)30147-0.pdf (paywalled)
    DOI 10.1016/j.tig.2016.10.008

    This paper might be even better: Beyond GWASs: Illuminating the Dark Road from Association to Function

    https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3824120/

    Abstract (emphasis mine):

    Genome-wide association studies (GWASs) have enabled the discovery of common genetic variation contributing to normal and pathological traits and clinical drug responses, but recognizing the precise targets of these associations is now the major challenge. Here, we review recent approaches to the functional follow-up of GWAS loci, including fine mapping of GWAS signal(s), prioritization of putative functional SNPs by the integration of genetic epidemiological and bioinformatic methods, and in vitro and in vivo experimental verification of predicted molecular mechanisms for identifying the targeted genes. The majority of GWAS-identified variants fall in noncoding regions of the genome. Therefore, this review focuses on strategies for assessing likely mechanisms affected by noncoding variants; such mechanisms include transcriptional regulation, noncoding RNA function, and epigenetic regulation. These approaches have already accelerated progress from genetic studies to biological knowledge and might ultimately guide the development of prognostic, preventive, and therapeutic measures.

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  70. @res

    No one denies individual differences in intelligence. This is strawmanning at its best (or worst).
     
    I specifically referred to denying the influence of genetics on intelligence. In other words, you strawmanned me by leaving out the "genetics" qualifier. But you are right that your first sentence was an example of strawmanning at its worst (that was what you meant, right?).

    As for no one denying the influence of genetics of on intelligence, if that is so then why did someone feel the need to write this article in 2017? Yes, There Is a Genetic Component to Intelligence https://www.thecut.com/2017/05/genetics-intelligence.html


    Given the plasticity of IQ and given the tremendous within race variation and between race overlap, you’d have to be pretty illiterate scientifically to claim that genes play a role in the black/white IQ gap. Moreover, a genetic interpretation wouldn’t necessarily support the conclusion you are pining for. If you followed the debate that was instigated by the Vox article, you’ll recall that Ezra Klein called James Flynn a few days before his encounter with Sam Harris. Flynn went on to say that it is entirely possible that the 10-point IQ gap reflects a 12 point environmental difference and a negative 2 point genetic difference. That is, blacks could very well have an inherent advantage in intelligence that is overwhelmed by environmental factors. This to me is more plausible than the alternative.
     
    That was some Turkheimer worthy obfuscation. Let's make this simple. Yes or no, do you believe there is exactly zero genetic contribution to the Black/White IQ difference in the US? Any answer other than an explicit "yes" indicates agreement with the conclusion I quoted from The Bell Curve. To repeat the most important part: "It seems highly likely to us that both genes and the environment have something to do with racial differences."

    On what basis do you think a black genetic advantage in intelligence is more likely than the opposite? I believe the opposite is more likely (has higher probability) because the required environmental difference is smaller if the phenotypic difference has the same sign as the genetic difference. A smaller required environmental difference is more likely.


    There’s nothing you can present that can’t be picked apart and comprehensively rebutted.
     
    The SNP frequencies differ between the races for the first (and likely almost all) IQ SNPs found. That is reality. Sorry you don't like it.

    You’re holding a house of cards, bruh. Even a baby’s breath can blown it down.

     

    Now that was some world class argumentation. I tremble before the awesomeness of your debating skills. /sarc

    To a rational, unbiased observer the second article served only to further bury Murray and his racist pseudoscience with much more detail than the first.
     
    Because the failure of the three authors to agree on a single response demonstrates just how compelling their arguments are. LOL!

    “the influence of genetics of on intelligence”

    Genes can’t influence intelligence because there are no psychophysical or psychological laws.

    Read More
    • Replies: @res
    Thank you for proving my point contra Okechukwu's comment 59.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  71. Tim too says:
    @APilgrim
    Most of the Purple Areas experienced STARVATION, during most of the 20th Century.

    During WWII, the entire Purple area was malnourished.

    North Koreans are still grossly malnourished. NORKS have TWICE the IQ of Blacks.

    well, the graphic in the post has no date on it, so how is anyone supposed to know? does it apply now? or during the mass starvations? WWII was a long time ago. Malnutrition in infancy/childhood does have lasting effects on intelligence. And who’s data are you going by for the DPRK malnutrition? And who’s IQ test is it? IQ depends on the test writer/creator as well as the test subjects. Remember, most people at select times past could not read, demographic IQ is not consistent over time. So the IQ for Russia eg, would be very different in say 1800, 1900, and 2000. That map of yours over time would show lots more red in many more places in the past.

    Over-interpretation of data is a problem. More refinement, definition/resolution of data is needed.

    Read More
    • Replies: @James Thompson
    As regards Russia, intelligence and literacy, some data here:

    https://www.unz.com/jthompson/50-russian-oblasts/
    , @APilgrim
    Do you have an ATF License for your rapid-fire, assault-question gun? Σωκρᾰ́της style excessive questions are NOT considered an argument, except @ law-schools & perhaps the Hemlock Society.

    I studied the 'Psychology of Individual Differences', & Advanced Mathematics at University, 50 years ago.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  72. Anon[378] • Disclaimer says:
    @Okechukwu

    Lynn and Vanhanen: Maroon is below 65, Purple is over 105: Deal with it.
     
    I'm familiar with the chart. What's shocking is that someone's dumb enough to believe it. Rather ironic in a discussion of intelligence.

    Funny how African athletes are able to compete in the normal Olympics rather than the Special Olympics.

    Mental retardation is cognitive limitation as characterized by scores greater than 2 standard deviations below the mean on a valid intelligence quotient (IQ) measure, with limitation of adaptive function in communication, self‐care, daily living skills at home or in the community, or social skills.

    https://www.sciencedirect.com/topics/neuroscience/mental-retardation

    Not only do Africans not have any of these issues, when they immigrate to the west they outperform whites.

    African immigrants are more educated than most — including people born in U.S.

    Batalova's research found that of the 1.4 million who are 25 and older, 41% have a bachelor's degree, compared with 30% of all immigrants and 32% of the U.S.-born population. Of the 19,000 U.S. immigrants from Norway — a country Trump reportedly told lawmakers is a good source of immigrants — 38% have college educations.

    The New American Economy study found that 1 in 3 of these undergraduate degrees were focused on science, technology, engineering and math — "training heavily in demand by today's employers."

    That report also found that African immigrants were significantly more likely to have graduate degrees. A total of 16% had a master's degree, medical degree, law degree or a doctorate, compared with 11% of the U.S.-born population.

    http://www.latimes.com/world/africa/la-fg-global-african-immigrants-explainer-20180112-story.html

    I don't know who's more stupid, Lynn and Vanhanen or people like you who accept their "research."

    > African immigrants are more educated than most — including people born in U.S.

    Lynn and Vanhanen showed the IQ in the native countries. You tried to counter that with specially selected elite African immigrants with that?? You do not seem to understand selection bias.

    You are a pathetic loser.

    Read More
    • Replies: @Okechukwu

    Lynn and Vanhanen showed the IQ in the native countries.
     
    Lynn and Vanhanen fabricated IQ data in the native countries. Unless you believe that a group of retarded children in Spain provide an accurate proxy for the people of Equatorial Guinea (among many other absurdities), then even a moron like you would have to admit that the Lynn/Vanhenen data is fraudulent.

    Actually, those pseudoscientific charlatans were counting precisely on ignorant people like you to receive their "research." But unfortunately for them, some real scientists got ahold of their data and proceeded to rip them new assholes. Beyond that, even laymen understand that Africans who speak multiple languages, function normally and perform every endeavor no matter how intellectually rigorous, CANNOT be functionally retarded. As a consequence, the only people that place any faith in the fraudulent Lynn/Vanhanen material are racist Internet white power trolls. For the rest of the global population, it's completely worthless and irrelevant. That's why you'll never see it quoted, cited or referenced by any credible source anywhere in the world. You see, outside of these echo-chambers and in the real world, we do this thing called due diligence. It's the thing that consigns all of these theories to the garbage heap.


    You tried to counter that with specially selected elite African immigrants with that?? You do not seem to understand selection bias.
     
    There is no selection bias in African immigration. Immigrants represent a cross-section of African society and not some elite. You know, it's really hard to get an elite via the visa lottery and chain migration. People who sponsor their relatives don't care how smart or dumb they are. Besides, basic common sense should tell you that you cannot get an elite capable of grossly outperforming whites from a source population of 55 IQ retards.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  73. Chainsaw1 says:
    @utu
    Keep in mind that it is not even possible to writes something like this "f(p)+f ‘(p)dx" in the case of the polygenic score. The independent variables are binary. One either have a given allele of SNP or not. The linearity in question concerns multiple variables not a single variable. The nonlinearity in terms of one variable does not even makes sense because variable is basically binary. If x and y are two different SNPs GWAS can approximate the answer to what is the difference in trait when x=0 or 1, i.e., when you have a given allele of the SNP or not. The same can be done for SNP y. But it is harder to answer whether the effect of x and y together is a sum of effects due to separate effects of x and of y. And when you have more and more SNPs that you want to include in the polygenic score it becomes harder and harder to disentangle their non liner interactions. In case of two binary variables x and y you can write polygenic score as: ax+by or ax+by+cxy. In the second case the nonlinear term xy, which kicks in only when both x and y are non zero, takes care of the possibility that the effect of two SNPs is not equal to the sum of the effects. This formulation makes it possible to account for the case when x and y individually have no effect (a=b=0) but in tandem they do. More or less this is how it should be done. The problem is that when you consider n SNPs in a linear case you have n variables but in nonlinear case with mixed terms you have additionally n*(n-1)/2 terms to consider. In the case of 1,000 SNPs you'll end up with 500,000 variables to be determined.

    > Keep in mind that it is not even possible to writes something like this “f(p)+f ‘(p)dx” in the case of the polygenic score. The independent variables are binary.

    Another stupid ignorant fool. Read the freaking supp before opening your big mouth. In PGS the independent variables are computed from the fraction distribution of the SNP variants, it can be as fine as you want with increasing sample size.

    1.5. Association Analyses

    Cohorts were asked to estimate this regression equation for each measured SNP:

    EduYears = Bo + B1*SNP + PC gamma + B*alpha + X + e, (1.1)

    where SNP is the allele dose of the SNP;

    https://www.biostars.org/p/75689/

    Assuming that you have a SNP: A/B and your genotype probablities are:

    A/A : 0.1
    A/B: 0.4
    B/B: 0.5

    (They should all sum to 1.0)

    Then the dosage for this SNP is: 0*A/A + 1*A/B + 2*B/B = 0.4 + 2*0.5 = 1.4

    So the maximum dosage you can get is 2.0 (that is if the genotype probabilies of 0 for A/A, A/B and 1.0 for B/B)

    With bigger sample size for determining the dosage, the finer the dosage value. Now shut up your stupid big mouth.

    Read More
    • Replies: @j2
    Yes, this is correct, thanks Chainsaw1. I tried to simplify it too much. The SNP frequency is the basic variable that in a large population becomes continuous, but in the polygenic score you take the weighted average of alleles (with the weights 0,1,2), called the dosage. It also can be considered continuous in a large population sample. The weights 0,1,2 do not accurately describe the contributions, but they do not need to. It is simply to create a continuous variable to compare to the effect in the measured variable (IQ). This is a simple and mathematically well justified method to measure effects of SNPs.

    So, the polygenic score, as a first order approximation, an additive variable, can reasonably well be used to compare close-by populations. But as the approximation is linear, it will not work well for very different populations, and one cannot estimate correctly the effect of the SNPs to genetically heritable IQ from the explanation power of the polygenic score in explaining the IQ variation. The real effect of these SNPs can be much larger, as the real effect is not linear but very probably caused by combinations of SNPs and is nowhere linear. This is not invalidating the polygenic score approach.

    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  74. res says:
    @RaceRealist88
    "the influence of genetics of on intelligence"

    Genes can't influence intelligence because there are no psychophysical or psychological laws.

    Thank you for proving my point contra Okechukwu’s comment 59.

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  75. utu says:

    Davide, is it you? When one day you succeed in constructing a predictor function of IQ it will be useful only if based on actual genotypes not the imputed ones. When it comes to a given SNP you either have no allele or have one allele or have two allies. You can scale and weight numbers 0,1,2 by whatever you want and it will not make any difference. Mathematically it will remain a discrete variable with three possible values. However, because the increment of dependent variable may not be the same when changing form to 0 to 1 allele as from 1 allele to 2 allies, in the predictor function instead of 0,1,2 as values of independent variable you need to use 0,1,Z, where Z needs to be determined ahead of time or you ca let the regression model circumvent this problem by using 2D nonlinear model and instead of coding one SNP with one variable you code it with two variables (x,y) that each can have value 0 or 1 and your model that you would fit to the trait would have three terms: Ax+By+Cxy, where A=B. This predictor function will have three possible values 0, A or C depending whether you have 0, 1 or 2 alleles. Now you see why it does not make much sense talking about “f(p)+f ‘(p)dx” because it is really a 2D problem and variables x and y are binary.

    Read More
    • Replies: @Chainsaw1
    > You can scale and weight numbers 0,1,2 by whatever you want and it will not make any difference. Mathematically it will remain a discrete variable with three possible values.

    You are really thick. Can you count beyond one?? Read the paper carefully. They are trying to fit the equation for a sample population, not individual. When you summed up for the sample you get the fractional values, something your two bit binary mind cannot digest. Stop wasting other people's time reading your crap.

    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  76. Chainsaw1 says:
    @utu
    Davide, is it you? When one day you succeed in constructing a predictor function of IQ it will be useful only if based on actual genotypes not the imputed ones. When it comes to a given SNP you either have no allele or have one allele or have two allies. You can scale and weight numbers 0,1,2 by whatever you want and it will not make any difference. Mathematically it will remain a discrete variable with three possible values. However, because the increment of dependent variable may not be the same when changing form to 0 to 1 allele as from 1 allele to 2 allies, in the predictor function instead of 0,1,2 as values of independent variable you need to use 0,1,Z, where Z needs to be determined ahead of time or you ca let the regression model circumvent this problem by using 2D nonlinear model and instead of coding one SNP with one variable you code it with two variables (x,y) that each can have value 0 or 1 and your model that you would fit to the trait would have three terms: Ax+By+Cxy, where A=B. This predictor function will have three possible values 0, A or C depending whether you have 0, 1 or 2 alleles. Now you see why it does not make much sense talking about “f(p)+f ‘(p)dx" because it is really a 2D problem and variables x and y are binary.

    > You can scale and weight numbers 0,1,2 by whatever you want and it will not make any difference. Mathematically it will remain a discrete variable with three possible values.

    You are really thick. Can you count beyond one?? Read the paper carefully. They are trying to fit the equation for a sample population, not individual. When you summed up for the sample you get the fractional values, something your two bit binary mind cannot digest. Stop wasting other people’s time reading your crap.

    Read More
    • Replies: @Johan Meyer
    Your bad language fails to hide your bad argument. I shall spell out your logical failure, to wit, why the issue remains when considering a population. Note that this is at most grade 12 probability.

    Let p_\ell be the frequency of allele \ell. Then q_\ell is the frequency of its absence. These probabilities sum to unity. There are four possibilities:

    Allele is doubly absent (probability is q_\ell^2).

    Allele is present in first carrying chromosomal homologue, but absent in second. Or allele is present in second, but missing in first. The probability of each is p_\ell q_\ell, for a total of 2p_\ell q_\ell. For low frequencies of the allele, this may be approximated as 2p_\ell, and for high frequencies, as 2q_\ell=2(1-p_\ell). Thus one may expect to find reasonable correlations away from middle grounds.

    Allele is present in both homologues (probability is p_\ell^2).

    It should not be too hard to redo the math replacing (p)robability with pq, i.e. p(1-p), and including a p^2 term with its own coefficient to be found.

    All of which is aside to whether the genes add directly to intelligence, or merely modulate the uptake of neurotoxins, which would produce the same twin correlation structure as would genes adding directly to IQ.

    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  77. j2 says:
    @Chainsaw1
    > Keep in mind that it is not even possible to writes something like this “f(p)+f ‘(p)dx” in the case of the polygenic score. The independent variables are binary.

    Another stupid ignorant fool. Read the freaking supp before opening your big mouth. In PGS the independent variables are computed from the fraction distribution of the SNP variants, it can be as fine as you want with increasing sample size.


    1.5. Association Analyses

    Cohorts were asked to estimate this regression equation for each measured SNP:

    EduYears = Bo + B1*SNP + PC gamma + B*alpha + X + e, (1.1)

    where SNP is the allele dose of the SNP;

     

    https://www.biostars.org/p/75689/

    Assuming that you have a SNP: A/B and your genotype probablities are:

    A/A : 0.1
    A/B: 0.4
    B/B: 0.5

    (They should all sum to 1.0)

    Then the dosage for this SNP is: 0*A/A + 1*A/B + 2*B/B = 0.4 + 2*0.5 = 1.4

    So the maximum dosage you can get is 2.0 (that is if the genotype probabilies of 0 for A/A, A/B and 1.0 for B/B)

     

    With bigger sample size for determining the dosage, the finer the dosage value. Now shut up your stupid big mouth.

    Yes, this is correct, thanks Chainsaw1. I tried to simplify it too much. The SNP frequency is the basic variable that in a large population becomes continuous, but in the polygenic score you take the weighted average of alleles (with the weights 0,1,2), called the dosage. It also can be considered continuous in a large population sample. The weights 0,1,2 do not accurately describe the contributions, but they do not need to. It is simply to create a continuous variable to compare to the effect in the measured variable (IQ). This is a simple and mathematically well justified method to measure effects of SNPs.

    So, the polygenic score, as a first order approximation, an additive variable, can reasonably well be used to compare close-by populations. But as the approximation is linear, it will not work well for very different populations, and one cannot estimate correctly the effect of the SNPs to genetically heritable IQ from the explanation power of the polygenic score in explaining the IQ variation. The real effect of these SNPs can be much larger, as the real effect is not linear but very probably caused by combinations of SNPs and is nowhere linear. This is not invalidating the polygenic score approach.

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  78. @Tim too
    well, the graphic in the post has no date on it, so how is anyone supposed to know? does it apply now? or during the mass starvations? WWII was a long time ago. Malnutrition in infancy/childhood does have lasting effects on intelligence. And who's data are you going by for the DPRK malnutrition? And who's IQ test is it? IQ depends on the test writer/creator as well as the test subjects. Remember, most people at select times past could not read, demographic IQ is not consistent over time. So the IQ for Russia eg, would be very different in say 1800, 1900, and 2000. That map of yours over time would show lots more red in many more places in the past.

    Over-interpretation of data is a problem. More refinement, definition/resolution of data is needed.

    As regards Russia, intelligence and literacy, some data here:

    https://www.unz.com/jthompson/50-russian-oblasts/

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  79. APilgrim says:

    Caucasian IQs display ‘Dolly-Parton’ curves.

    A bimodal distribution is a continuous probability distribution with two different modes. These appear as distinct peaks (local maxima) in the probability density function. Ashkenazim IQ distribution has a small, but statistically significant ‘spike’, in the extreme high range.

    Dolly Parton joked that she was an early joiner of the ‘Burn the Bra Movement’ & it took the Fire Department 4 days to put it out.

    Read More
    • Replies: @Factorize
    A normally distributed IQ distribution follows as a logical consequence of having a very very large number of small effect common SNPs. This extreme level of polygenicity is the driving force that shapes the distribution; this is more important than even SES segregation as was seen in a recent study. In that study it was shown that the highly polygenic nature of IQ resulted in ongoing reshuffling of PGS and SES between generations. The elitist view of permanent social stratification is not supported by current evidence.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  80. APilgrim says:
    @Tim too
    well, the graphic in the post has no date on it, so how is anyone supposed to know? does it apply now? or during the mass starvations? WWII was a long time ago. Malnutrition in infancy/childhood does have lasting effects on intelligence. And who's data are you going by for the DPRK malnutrition? And who's IQ test is it? IQ depends on the test writer/creator as well as the test subjects. Remember, most people at select times past could not read, demographic IQ is not consistent over time. So the IQ for Russia eg, would be very different in say 1800, 1900, and 2000. That map of yours over time would show lots more red in many more places in the past.

    Over-interpretation of data is a problem. More refinement, definition/resolution of data is needed.

    Do you have an ATF License for your rapid-fire, assault-question gun? Σωκρᾰ́της style excessive questions are NOT considered an argument, except @ law-schools & perhaps the Hemlock Society.

    I studied the ‘Psychology of Individual Differences’, & Advanced Mathematics at University, 50 years ago.

    Read More
    • Replies: @Tim too
    thank you APilgrim. I'm not questioning that your chart means something, I'm just not sure what it means. But if your point is that genetics is involved, I agree, I do think genetics is involved. Along with other things.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  81. @Chainsaw1
    > You can scale and weight numbers 0,1,2 by whatever you want and it will not make any difference. Mathematically it will remain a discrete variable with three possible values.

    You are really thick. Can you count beyond one?? Read the paper carefully. They are trying to fit the equation for a sample population, not individual. When you summed up for the sample you get the fractional values, something your two bit binary mind cannot digest. Stop wasting other people's time reading your crap.

    Your bad language fails to hide your bad argument. I shall spell out your logical failure, to wit, why the issue remains when considering a population. Note that this is at most grade 12 probability.

    Let p_\ell be the frequency of allele \ell. Then q_\ell is the frequency of its absence. These probabilities sum to unity. There are four possibilities:

    Allele is doubly absent (probability is q_\ell^2).

    Allele is present in first carrying chromosomal homologue, but absent in second. Or allele is present in second, but missing in first. The probability of each is p_\ell q_\ell, for a total of 2p_\ell q_\ell. For low frequencies of the allele, this may be approximated as 2p_\ell, and for high frequencies, as 2q_\ell=2(1-p_\ell). Thus one may expect to find reasonable correlations away from middle grounds.

    Allele is present in both homologues (probability is p_\ell^2).

    It should not be too hard to redo the math replacing (p)robability with pq, i.e. p(1-p), and including a p^2 term with its own coefficient to be found.

    All of which is aside to whether the genes add directly to intelligence, or merely modulate the uptake of neurotoxins, which would produce the same twin correlation structure as would genes adding directly to IQ.

    Read More
    • Replies: @Johan Meyer
    One defect in the above is that I use p as the probability that a given homologue of the carrying chromosome will carry the allele, yet conflate it with the population frequency. This (p) is half of the population frequency of the allele, as the maximum frequency is 200% (everyone has two copies). For the math in that comment to work, use

    p=population frequency/2
    , @j2
    Ok, Johan. I agree with your comment about Chansaw1's style of presenting his arguments, but let us not go to this, especially as you also made a mistake right there. You have assumed that the population is in Harvey-Weinstein equilibrium, which it does not necessarily need to be, thus it is not always xx, 2x(1-x), (1-x)(1-x). Add the assumption and then it is fine.
    , @utu
    The frequencies f0,f1,f2, where f0+f1+f2=1 for any allele in any population that indicate frequency (probability) of occurrence of 0 allele, of 1 allele and of 2 allies, respectively are empirical values for a given sample. These frequencies suffice to determine the first moment of the distribution of the predicted trait by a polygenic score. But to determine the 2nd moment the frequencies alone are insufficient and a covariance matrix of SNPs included in the polygenic score must be used. The distribution is determined by a sample of N individuals. The polygenic score if only implicitly must be spelled out for each individual and in this score the frequencies do not play any role.
    , @Chainsaw1
    You are unnecessary complicate thing by putting in all that extra variable and subscript, using procedures that do not conform to the format of the data available. In PGS effective allele frequncy (EAF) is the frequency of the effective allele in the dataset and it is already given in decimal fraction frequency, e.g.

    SNP,A1,A2,EAF,Beta,SE,Pval
    rs9859556,T,G,0.6905,0.029,0.001,3.98E-91

    where A1 is the effective allele and Beta is the effect size for the sample. They do not care about q which has zero, zilch, nada direct effect on the PGS score. And they are not interested in doing algebraic manipulation, what they want is to have minimum prediction error over the whole sample. So all your algebraic manipulations are for freaking zilch, nothing.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  82. @Johan Meyer
    Your bad language fails to hide your bad argument. I shall spell out your logical failure, to wit, why the issue remains when considering a population. Note that this is at most grade 12 probability.

    Let p_\ell be the frequency of allele \ell. Then q_\ell is the frequency of its absence. These probabilities sum to unity. There are four possibilities:

    Allele is doubly absent (probability is q_\ell^2).

    Allele is present in first carrying chromosomal homologue, but absent in second. Or allele is present in second, but missing in first. The probability of each is p_\ell q_\ell, for a total of 2p_\ell q_\ell. For low frequencies of the allele, this may be approximated as 2p_\ell, and for high frequencies, as 2q_\ell=2(1-p_\ell). Thus one may expect to find reasonable correlations away from middle grounds.

    Allele is present in both homologues (probability is p_\ell^2).

    It should not be too hard to redo the math replacing (p)robability with pq, i.e. p(1-p), and including a p^2 term with its own coefficient to be found.

    All of which is aside to whether the genes add directly to intelligence, or merely modulate the uptake of neurotoxins, which would produce the same twin correlation structure as would genes adding directly to IQ.

    One defect in the above is that I use p as the probability that a given homologue of the carrying chromosome will carry the allele, yet conflate it with the population frequency. This (p) is half of the population frequency of the allele, as the maximum frequency is 200% (everyone has two copies). For the math in that comment to work, use

    p=population frequency/2

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  83. j2 says:
    @Johan Meyer
    Your bad language fails to hide your bad argument. I shall spell out your logical failure, to wit, why the issue remains when considering a population. Note that this is at most grade 12 probability.

    Let p_\ell be the frequency of allele \ell. Then q_\ell is the frequency of its absence. These probabilities sum to unity. There are four possibilities:

    Allele is doubly absent (probability is q_\ell^2).

    Allele is present in first carrying chromosomal homologue, but absent in second. Or allele is present in second, but missing in first. The probability of each is p_\ell q_\ell, for a total of 2p_\ell q_\ell. For low frequencies of the allele, this may be approximated as 2p_\ell, and for high frequencies, as 2q_\ell=2(1-p_\ell). Thus one may expect to find reasonable correlations away from middle grounds.

    Allele is present in both homologues (probability is p_\ell^2).

    It should not be too hard to redo the math replacing (p)robability with pq, i.e. p(1-p), and including a p^2 term with its own coefficient to be found.

    All of which is aside to whether the genes add directly to intelligence, or merely modulate the uptake of neurotoxins, which would produce the same twin correlation structure as would genes adding directly to IQ.

    Ok, Johan. I agree with your comment about Chansaw1′s style of presenting his arguments, but let us not go to this, especially as you also made a mistake right there. You have assumed that the population is in Harvey-Weinstein equilibrium, which it does not necessarily need to be, thus it is not always xx, 2x(1-x), (1-x)(1-x). Add the assumption and then it is fine.

    Read More
    • Replies: @Johan Meyer
    Hardy Weinberg, though point taken.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  84. APilgrim says:

    WRT the Global IQ distribution map above … Asian IQ Averages (105+) & Black African IQ Averages (65-) each follow normal distribution curves.

    Caucasian IQs are bimodal, with local maximas on the order of ~90 and ~110. Ashkenazim IQs are closer to a normal distribution, with a noticeable ‘spike’ @ ~150. So, qualitatively, there are a higher percentage of extreme intelligence among Ashkenazim than among other Caucasians, and/or Asians.

    ‘Average’ (Mean, Mode) IQ for Ashkenazim & other Caucasians is less useful, than for Blacks & Asians, due to the divergence from Normal distribution curves.

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter Display All Comments
  85. @j2
    Ok, Johan. I agree with your comment about Chansaw1's style of presenting his arguments, but let us not go to this, especially as you also made a mistake right there. You have assumed that the population is in Harvey-Weinstein equilibrium, which it does not necessarily need to be, thus it is not always xx, 2x(1-x), (1-x)(1-x). Add the assumption and then it is fine.

    Hardy Weinberg, though point taken.

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  86. Factorize says:
    @APilgrim
    Caucasian IQs display 'Dolly-Parton' curves.

    A bimodal distribution is a continuous probability distribution with two different modes. These appear as distinct peaks (local maxima) in the probability density function. Ashkenazim IQ distribution has a small, but statistically significant 'spike', in the extreme high range.

    https://upload.wikimedia.org/wikipedia/commons/e/e2/Bimodal.png

    Dolly Parton joked that she was an early joiner of the 'Burn the Bra Movement' & it took the Fire Department 4 days to put it out.

    A normally distributed IQ distribution follows as a logical consequence of having a very very large number of small effect common SNPs. This extreme level of polygenicity is the driving force that shapes the distribution; this is more important than even SES segregation as was seen in a recent study. In that study it was shown that the highly polygenic nature of IQ resulted in ongoing reshuffling of PGS and SES between generations. The elitist view of permanent social stratification is not supported by current evidence.

    Read More
    • Replies: @APilgrim
    Caucasian IQ distribution generally, and Ashkenazim IQ distribution particularly do NOT follow a smooth normal distribution curve. No increase in sample size will alter that reality.

    Quantum effects, such as radioactive decay, photon emission ... etc do not generally produce a normal smooth distribution. The emissions are discrete, with probabilistic occurrences.

    The universe is real, not ideal. Deal with it.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  87. APilgrim says:
    @Factorize
    A normally distributed IQ distribution follows as a logical consequence of having a very very large number of small effect common SNPs. This extreme level of polygenicity is the driving force that shapes the distribution; this is more important than even SES segregation as was seen in a recent study. In that study it was shown that the highly polygenic nature of IQ resulted in ongoing reshuffling of PGS and SES between generations. The elitist view of permanent social stratification is not supported by current evidence.

    Caucasian IQ distribution generally, and Ashkenazim IQ distribution particularly do NOT follow a smooth normal distribution curve. No increase in sample size will alter that reality.

    Quantum effects, such as radioactive decay, photon emission … etc do not generally produce a normal smooth distribution. The emissions are discrete, with probabilistic occurrences.

    The universe is real, not ideal. Deal with it.

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  88. Factorize says:

    I can roll with reality no matter how it zig or zags, though considering the central limit theorem and that there are possibly tens of thousands of SNPs of tiny effect size, the inescapable conclusion is that a normal distribution describes IQ. We have N>>30 so we can safely assume normality. It would be very easy to run a simulation to prove this. Any variance from normality would have to be relatively small (depends on how many large rare effect SNPs might be out there). If normality did not apply, then reporting IQ in SD based on the assumption of normality would not be accurate.

    Read More
    • Replies: @res
    Having multiple sub-populations (e.g. Ashkenazi Jews and various other groups of "whites") is likely to break the genetic normality. However, depending on the group and techniques used to norm the tests, the IQ test results might still be normal.

    There is also a likely departure from genetic normality because large effect negative mutations (and other genetic issues like Trisomy 21, aka Down Syndrome) are more common than the positive kind.

    The normal model is not reality, but it is a fairly good approximation in this case.

    I'm not sure which is more frustrating:
    "Reality exactly conforms to my model."
    "A small difference between the model and reality means the model is worthless."

    But they certainly provide nice reference points for the false dichotomy.
    , @j2
    IQ is measured by a set of questions. The scores are fitted by a standardization cohort to a normal distribution, because how else you can know how difficult individual questions were and what score a person with a given intelligence should get. So, when scaled in this way, the tails are not any thicker or thinner than they should be and if the standardization cohort is correctly chose, the curve has a bell shape and gives quite the frequencies to each IQ level as it should. As heavily mentally retarded are usually excluded, the low part is cut at some point and continued by a special way, which is also normalized to give normally distributed values, and for super-high IQ you need to do the same thing.

    About US Ashkenazi IQ I looked some plots that Razim Khan had measured with Wordsum. Their IQ distribution was not normal and the average was 107.5 about. It was skewed like if two generations ago there was cutting of the below 100 part very small by selective migration and then the distribution smoothened at the cut-point. The other distributions looked normal in those plots. There were no Dolly Partons for White non-Jewish Americans. Maybe the Dolly Parton is for white latinos and white europeans.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  89. res says:
    @Factorize
    I can roll with reality no matter how it zig or zags, though considering the central limit theorem and that there are possibly tens of thousands of SNPs of tiny effect size, the inescapable conclusion is that a normal distribution describes IQ. We have N>>30 so we can safely assume normality. It would be very easy to run a simulation to prove this. Any variance from normality would have to be relatively small (depends on how many large rare effect SNPs might be out there). If normality did not apply, then reporting IQ in SD based on the assumption of normality would not be accurate.

    Having multiple sub-populations (e.g. Ashkenazi Jews and various other groups of “whites”) is likely to break the genetic normality. However, depending on the group and techniques used to norm the tests, the IQ test results might still be normal.

    There is also a likely departure from genetic normality because large effect negative mutations (and other genetic issues like Trisomy 21, aka Down Syndrome) are more common than the positive kind.

    The normal model is not reality, but it is a fairly good approximation in this case.

    I’m not sure which is more frustrating:
    “Reality exactly conforms to my model.”
    “A small difference between the model and reality means the model is worthless.”

    But they certainly provide nice reference points for the false dichotomy.

    Read More
    • Replies: @APilgrim
    Sorry that your probability & statistics survey course did not get to the back of the book. The differences in distribution samples can be BIG & IMPORTANT, as in IQ distributions.

    Multimodal distribution is a distribution that has multiple modes (thus two or more "peaks"). Multimodality of the distribution in a sample is often a strong indication that the distribution of the variable in population is not normal. Multimodality of the distribution may provide important information about the nature of the investigated variable (i.e., the measured quality). https://link.springer.com/article/10.1007/BF01029273

    http://documentation.statsoft.com/STATISTICAHelp/Glossary/Images/MULTIMOD.GIF

    Deal with it.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  90. j2 says:
    @Factorize
    I can roll with reality no matter how it zig or zags, though considering the central limit theorem and that there are possibly tens of thousands of SNPs of tiny effect size, the inescapable conclusion is that a normal distribution describes IQ. We have N>>30 so we can safely assume normality. It would be very easy to run a simulation to prove this. Any variance from normality would have to be relatively small (depends on how many large rare effect SNPs might be out there). If normality did not apply, then reporting IQ in SD based on the assumption of normality would not be accurate.

    IQ is measured by a set of questions. The scores are fitted by a standardization cohort to a normal distribution, because how else you can know how difficult individual questions were and what score a person with a given intelligence should get. So, when scaled in this way, the tails are not any thicker or thinner than they should be and if the standardization cohort is correctly chose, the curve has a bell shape and gives quite the frequencies to each IQ level as it should. As heavily mentally retarded are usually excluded, the low part is cut at some point and continued by a special way, which is also normalized to give normally distributed values, and for super-high IQ you need to do the same thing.

    About US Ashkenazi IQ I looked some plots that Razim Khan had measured with Wordsum. Their IQ distribution was not normal and the average was 107.5 about. It was skewed like if two generations ago there was cutting of the below 100 part very small by selective migration and then the distribution smoothened at the cut-point. The other distributions looked normal in those plots. There were no Dolly Partons for White non-Jewish Americans. Maybe the Dolly Parton is for white latinos and white europeans.

    Read More
    • Replies: @Factorize
    j2, the normal distribution for IQ arose from empirical observations. Pschometricians looked out in the community and saw that most people were of average intelligence and some were of greater and some were of lower intelligence. A normal distribution results when you add up many random variables as with IQ. It is the logical outcome for the genetic architecture of intelligence that has been found.

    The idea of a multimodal distribution is not supported by the evidence. Recent research has found that different socioeconomic groups are not that exceptionally different. Reshuffling of the SNPs every generation tends to pull everyone back to average.

    Life would be structured much differently if this were not true. If there were a multimodal IQ distribution it would have been much more necessary to have separate schools for those of differing ability levels. This was by and large not found to be necessary.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  91. APilgrim says:
    @res
    Having multiple sub-populations (e.g. Ashkenazi Jews and various other groups of "whites") is likely to break the genetic normality. However, depending on the group and techniques used to norm the tests, the IQ test results might still be normal.

    There is also a likely departure from genetic normality because large effect negative mutations (and other genetic issues like Trisomy 21, aka Down Syndrome) are more common than the positive kind.

    The normal model is not reality, but it is a fairly good approximation in this case.

    I'm not sure which is more frustrating:
    "Reality exactly conforms to my model."
    "A small difference between the model and reality means the model is worthless."

    But they certainly provide nice reference points for the false dichotomy.

    Sorry that your probability & statistics survey course did not get to the back of the book. The differences in distribution samples can be BIG & IMPORTANT, as in IQ distributions.

    Multimodal distribution is a distribution that has multiple modes (thus two or more “peaks”). Multimodality of the distribution in a sample is often a strong indication that the distribution of the variable in population is not normal. Multimodality of the distribution may provide important information about the nature of the investigated variable (i.e., the measured quality). https://link.springer.com/article/10.1007/BF01029273

    Deal with it.

    Read More
    • Replies: @Factorize
    This is not worth argung about: the distribution of PGS scores should be almost nearly perfectly normal. It would be no great effort to run a sinulation with the 1271 SNPs to demonstrate this. The twin peaks model is not valid. I would be surprised in a sample of this massive scale whether it would deviate by even 1 part per thousand from normality. Considering that this is social science and not mathematical science, one part per thousand error is not bad. If you were to use only the SNPs in the PGS even those with Down's syndrome or severe learning disabilities would likely reproduce a normal distribution. While white European might have become a somewhat ambiguous classifier, one would expect that this still has some genetic meaning.
    , @res
    Yes, because the IQ curve looks like that in reality. Do you think the IQ curve in the US is multimodal? If so, evidence? It is certainly not multimodal to the extreme of the strawman graphic you present.

    I agree that there are some distortions to normality (that was pretty much the point of my comment, if you read the whole thing, though I would have thought you would at least have noticed the very first sentence!) in the US IQ distribution due to different subpopulations. The major effects I see from that are:
    - Disproportionate representation for various groups at both extremes of the distribution.
    - Possible fat far right tail due to high IQ subpopulations. Which may be quite important if extremely high IQ is important for innovation (which I happen to think is the case).

    As far as I can tell, within +-2SD (roughly 95% of people) of the mean the normal distribution provides a pretty good model for the overall US IQ distribution.

    That you failed to understand the nuances of my full comment says much more about your reading comprehension than it does about my probability and statistics background.

    P.S. I take it you are staking out the "A small difference between the model and reality means the model is worthless" end of the false dichotomy.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  92. hyperbola says:
    @utu

    These studies and their “canned” forms of analysis are wearing out my patience to keep reading them!
     
    You should look at them and try to unpack them. There are legitimate questions. Since twin studies show that IQ score is highly heritable then a predictor function of IQ score based on SNPs should be possible to construct. It is possible as some critics say that the twin studies base heritability is overestimated but I think nobody believes that it should be zero. Therefore a nonzero function mapping genes onto the scale of IQ scores must exist. Because mathematically the problem is insanely undetermined (7 billion people and almost infinite number of combinations out of 10 million SNP set) the problem is not finding such a function but finding the one that is not spurious as the result of overfitting.

    I have read somewhere that 88% GWAS hits usually come from the noncoding part of DNA in studies of all kinds of different traits. So what is the biological mechanism how noncoding SPNs affect a trait? This is an importnt problem which however should not stop us from looking for mathematically constructed functions even if it includes SNPs that biologists do not know what is their role.

    IQ score, intelligence, cognitive abilities , educational attainments pose additional problem as they all are related to some not well defined trait which certainly is very complex. It is not like height that has a good definition and can be measured with a physical scale. Nevertheless twin studies suggest that heritability within our type of society is around 50-70%. Obviously if we could include twins raised by wolves the result would be much different.

    As far as reading the papers it is really hard because there is way too much of unnecessary professional jargon that obfuscates even if not intentionally, though I have some doubts about how unintentional it is. Look at this paper and you will see how scrupulous they are about reporting P-values of anything and everything. This is mostly BS. But on the other hand when it comes to spell out the polygenic score definition that contains 1,000,000 SPNs they are silent. They have mentioned it en passant in the FAQ section only. Or look at a brief paragraph when they used the predictor (white) function on Afro-American population. They do not explain which predictor function they used. Was it the one that produced 3.9%. R^2 or the one that produced 11-13% R^2 on white population? They also could have told us what was the offset bias between the two populations.

    I would not mind if there was a higher authority that could exercise its power of subpoena over the cliques of scientist and have them undergo interrogations where they would have to give some explanations. I could volunteer for the position of Torquemada.

    I have read somewhere that 88% GWAS hits usually come from the noncoding part of DNA in studies of all kinds of different traits. So what is the biological mechanism how noncoding SPNs affect a trait?

    For humans the estimates seem to be that protein-coding regions encompass on the order of 1% of DNA. There are other regions that are used to produce various important RNAs such as tRNA, microRNA, etc. Still other regions seem to be at least transcribed although no known (direct) functions are known and they may not be conserved evolutionarily. Other regions may show epigenetic effects (DNA methylation), but without identification of a specific functional activity. This is a very complex subject – you might find this useful.

    My guess would be that many of the “junk” regions have to do with the spatial structure and packaging of DNA and hence on expression of different protein-coding regions.

    Defining functional DNA elements in the human genome

    https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4035993/

    As for 10 million SNPs, this study only measured 1 million people. Can we expect 7000 times more SNPs when we expand to 7 billion humans? We already know that many of the SNPs seem to be in DNA regions that are not under evolutionary selection (see paper).

    Read More
    • Replies: @utu
    I am looking at the problem form the mathematical end. I am not really interested in mechanism what SNPs do or not. It appears that nobody really does know what exactly SPNs form the noncoding regions do but everybody agrees that they are important because they must be included in polygenic scores to give the scores predictive power.

    I do not know how phone numbers associated with different people made those people to move to particular geographic locations or was it the others way around. I do not care what is the direction of the causality vector. But having a sufficiently large sample of people, their phone numbers and their geographic locations I can discover the rule of this association. And if I can confirm the rule on an independent validation sample I will be able to decode (lat,long) from phone numbers.

    This is really the bottom line of IQ nature-nurture controversy. What is the maximal prediction power with any polygenic score? Is it close to heritability based on twin studies or not. How this or that SNP make somebody smarter or not is secondary to me.

    As for 10 million SNPs, this study only measured 1 million people. Can we expect 7000 times more SNPs when we expand to 7 billion humans?
     
    What are you talking about? Things do not scale like that.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  93. hyperbola says:
    @j2
    "So what is the biological mechanism how noncoding SPNs affect a trait?"

    I guess it usually is a noncoding part. Coding parts code different proteins, which may change something but normally to the worse. Non-coding part contains control parts that modify how long some protein is produced. So, for instance, the brain gets bigger if the growing stage (when some proteins are produced) is longer. Non-coding is not the same as junk DNA.

    “So what is the biological mechanism how noncoding SPNs affect a trait?”

    I guess it usually is a noncoding part…..

    A very complex subject.

    Defining functional DNA elements in the human genome

    https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4035993/

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  94. Factorize says:
    @APilgrim
    Sorry that your probability & statistics survey course did not get to the back of the book. The differences in distribution samples can be BIG & IMPORTANT, as in IQ distributions.

    Multimodal distribution is a distribution that has multiple modes (thus two or more "peaks"). Multimodality of the distribution in a sample is often a strong indication that the distribution of the variable in population is not normal. Multimodality of the distribution may provide important information about the nature of the investigated variable (i.e., the measured quality). https://link.springer.com/article/10.1007/BF01029273

    http://documentation.statsoft.com/STATISTICAHelp/Glossary/Images/MULTIMOD.GIF

    Deal with it.

    This is not worth argung about: the distribution of PGS scores should be almost nearly perfectly normal. It would be no great effort to run a sinulation with the 1271 SNPs to demonstrate this. The twin peaks model is not valid. I would be surprised in a sample of this massive scale whether it would deviate by even 1 part per thousand from normality. Considering that this is social science and not mathematical science, one part per thousand error is not bad. If you were to use only the SNPs in the PGS even those with Down’s syndrome or severe learning disabilities would likely reproduce a normal distribution. While white European might have become a somewhat ambiguous classifier, one would expect that this still has some genetic meaning.

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  95. Factorize says:
    @j2
    IQ is measured by a set of questions. The scores are fitted by a standardization cohort to a normal distribution, because how else you can know how difficult individual questions were and what score a person with a given intelligence should get. So, when scaled in this way, the tails are not any thicker or thinner than they should be and if the standardization cohort is correctly chose, the curve has a bell shape and gives quite the frequencies to each IQ level as it should. As heavily mentally retarded are usually excluded, the low part is cut at some point and continued by a special way, which is also normalized to give normally distributed values, and for super-high IQ you need to do the same thing.

    About US Ashkenazi IQ I looked some plots that Razim Khan had measured with Wordsum. Their IQ distribution was not normal and the average was 107.5 about. It was skewed like if two generations ago there was cutting of the below 100 part very small by selective migration and then the distribution smoothened at the cut-point. The other distributions looked normal in those plots. There were no Dolly Partons for White non-Jewish Americans. Maybe the Dolly Parton is for white latinos and white europeans.

    j2, the normal distribution for IQ arose from empirical observations. Pschometricians looked out in the community and saw that most people were of average intelligence and some were of greater and some were of lower intelligence. A normal distribution results when you add up many random variables as with IQ. It is the logical outcome for the genetic architecture of intelligence that has been found.

    The idea of a multimodal distribution is not supported by the evidence. Recent research has found that different socioeconomic groups are not that exceptionally different. Reshuffling of the SNPs every generation tends to pull everyone back to average.

    Life would be structured much differently if this were not true. If there were a multimodal IQ distribution it would have been much more necessary to have separate schools for those of differing ability levels. This was by and large not found to be necessary.

    Read More
    • Replies: @utu

    the normal distribution for IQ arose from empirical observations
     
    It appears to be so but do not forget that the IQ scale is kind of arbitrary. It was constructed. And I would not be surprised if obtaining Gaussian distribution was the objective of the engineers.

    If your n-th question in your test is answered by only 1.5% of population you will know what weight to assign to that question in the final score to be on the track to Gaussian distribution construction.

    And in the end you can always perform a nonlinear scale adjustment. Any concave pfd(IQ) distribution can be transformed into a Gaussian one by some monotonic f(IQ)-->IQ transformation.

    The bottom line is that whether IQ score is Gaussian or not is unimportant. But on the other hand what kind of distributions can be constructed using a polygenic score might be more interesting.
    , @j2
    Look, Factorize, safely assume I know all about basics of statistics or any basic math and have proven all those theorems in exercises ages ago. We know that we get the normal distribution under weak conditions, but if you have an IQ test with a set of questions and the distribution is not normal, you assume that, sorry, it has to be normal as it is normal under weak conditions, so my test set is just poorly chosen. Like, make an IQ test set with half trivially simple questions and half unbelievably difficult questions, you get the distribution that almost all got only the trivially simple questions, but one guy (who stole the test questions in advance) got everything correct. This is empirical, and you conclude that empirical is not correct, the test set was poorly chosen, if correctly chosen it must give normal. And this is not because of engineers, they do not know anything for sure, it is mathematicians, they have all the proofs.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  96. utu says:
    @Johan Meyer
    Your bad language fails to hide your bad argument. I shall spell out your logical failure, to wit, why the issue remains when considering a population. Note that this is at most grade 12 probability.

    Let p_\ell be the frequency of allele \ell. Then q_\ell is the frequency of its absence. These probabilities sum to unity. There are four possibilities:

    Allele is doubly absent (probability is q_\ell^2).

    Allele is present in first carrying chromosomal homologue, but absent in second. Or allele is present in second, but missing in first. The probability of each is p_\ell q_\ell, for a total of 2p_\ell q_\ell. For low frequencies of the allele, this may be approximated as 2p_\ell, and for high frequencies, as 2q_\ell=2(1-p_\ell). Thus one may expect to find reasonable correlations away from middle grounds.

    Allele is present in both homologues (probability is p_\ell^2).

    It should not be too hard to redo the math replacing (p)robability with pq, i.e. p(1-p), and including a p^2 term with its own coefficient to be found.

    All of which is aside to whether the genes add directly to intelligence, or merely modulate the uptake of neurotoxins, which would produce the same twin correlation structure as would genes adding directly to IQ.

    The frequencies f0,f1,f2, where f0+f1+f2=1 for any allele in any population that indicate frequency (probability) of occurrence of 0 allele, of 1 allele and of 2 allies, respectively are empirical values for a given sample. These frequencies suffice to determine the first moment of the distribution of the predicted trait by a polygenic score. But to determine the 2nd moment the frequencies alone are insufficient and a covariance matrix of SNPs included in the polygenic score must be used. The distribution is determined by a sample of N individuals. The polygenic score if only implicitly must be spelled out for each individual and in this score the frequencies do not play any role.

    Read More
    • Replies: @Johan Meyer
    If you want to connect genetically (directly caused) individual IQ, as a random variable, to (genetically caused) group IQ, then such a more careful calculation is needed to check the correctness of the result obtained from using only the means and frequencies of genes.

    This issue is dealt with by handwaving about units of aggregation, without offering hypotheses as to what may account for discrepancies between levels of aggregation. (The link is from the HBD reading list).

    I do not per se object to such aggregation, and it is unavoidable in studies of low frequency systems e.g. crime, whether one takes genetic, environmental or combined approaches.

    If one's aim is to check whether group means can be predicted based on genes, without positing a hypothesis about individual scores, the approach is fine, as long as one does not pretend that one has demonstrated things that were not demonstrated.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  97. res says:
    @APilgrim
    Sorry that your probability & statistics survey course did not get to the back of the book. The differences in distribution samples can be BIG & IMPORTANT, as in IQ distributions.

    Multimodal distribution is a distribution that has multiple modes (thus two or more "peaks"). Multimodality of the distribution in a sample is often a strong indication that the distribution of the variable in population is not normal. Multimodality of the distribution may provide important information about the nature of the investigated variable (i.e., the measured quality). https://link.springer.com/article/10.1007/BF01029273

    http://documentation.statsoft.com/STATISTICAHelp/Glossary/Images/MULTIMOD.GIF

    Deal with it.

    Yes, because the IQ curve looks like that in reality. Do you think the IQ curve in the US is multimodal? If so, evidence? It is certainly not multimodal to the extreme of the strawman graphic you present.

    I agree that there are some distortions to normality (that was pretty much the point of my comment, if you read the whole thing, though I would have thought you would at least have noticed the very first sentence!) in the US IQ distribution due to different subpopulations. The major effects I see from that are:
    - Disproportionate representation for various groups at both extremes of the distribution.
    - Possible fat far right tail due to high IQ subpopulations. Which may be quite important if extremely high IQ is important for innovation (which I happen to think is the case).

    As far as I can tell, within +-2SD (roughly 95% of people) of the mean the normal distribution provides a pretty good model for the overall US IQ distribution.

    That you failed to understand the nuances of my full comment says much more about your reading comprehension than it does about my probability and statistics background.

    P.S. I take it you are staking out the “A small difference between the model and reality means the model is worthless” end of the false dichotomy.

    Read More
    • Replies: @APilgrim
    You take it wrong. Models serve constructive purposes, as do graphics. I routinely start with a simple model and/or engineering approximation. Then advance from heuristics to qualitative math & more advanced quantitative models. I may carry solutions to a closed form exact answer, which can be exactly modeled, and experimentally verified, if the project calls for it.

    American Caucasians, of European extraction, had a pronounced bimodal distribution, from the beginning of the IQ studies, until the late 1970s. At that point, most of these subjects, were banned by the Political Correct (PC) 'thought-police'. IQ Tests and similar protocols such as SAT, ACT, MCAT, LSAT and the like were widely used to 'weed-out', the stupid white kids, from top academic programs. These tests clearly favored the bright kids among Jews & Asians. They also virtually eliminated Blacks from consideration. Low IQ Blacks have been shoe-horned into lots of academic programs, for which they are not qualified.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  98. Tim too says:
    @APilgrim
    Do you have an ATF License for your rapid-fire, assault-question gun? Σωκρᾰ́της style excessive questions are NOT considered an argument, except @ law-schools & perhaps the Hemlock Society.

    I studied the 'Psychology of Individual Differences', & Advanced Mathematics at University, 50 years ago.

    thank you APilgrim. I’m not questioning that your chart means something, I’m just not sure what it means. But if your point is that genetics is involved, I agree, I do think genetics is involved. Along with other things.

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  99. utu says:
    @Factorize
    j2, the normal distribution for IQ arose from empirical observations. Pschometricians looked out in the community and saw that most people were of average intelligence and some were of greater and some were of lower intelligence. A normal distribution results when you add up many random variables as with IQ. It is the logical outcome for the genetic architecture of intelligence that has been found.

    The idea of a multimodal distribution is not supported by the evidence. Recent research has found that different socioeconomic groups are not that exceptionally different. Reshuffling of the SNPs every generation tends to pull everyone back to average.

    Life would be structured much differently if this were not true. If there were a multimodal IQ distribution it would have been much more necessary to have separate schools for those of differing ability levels. This was by and large not found to be necessary.

    the normal distribution for IQ arose from empirical observations

    It appears to be so but do not forget that the IQ scale is kind of arbitrary. It was constructed. And I would not be surprised if obtaining Gaussian distribution was the objective of the engineers.

    If your n-th question in your test is answered by only 1.5% of population you will know what weight to assign to that question in the final score to be on the track to Gaussian distribution construction.

    And in the end you can always perform a nonlinear scale adjustment. Any concave pfd(IQ) distribution can be transformed into a Gaussian one by some monotonic f(IQ)–>IQ transformation.

    The bottom line is that whether IQ score is Gaussian or not is unimportant. But on the other hand what kind of distributions can be constructed using a polygenic score might be more interesting.

    Read More
    • Replies: @Factorize
    utu, I have been considering your comments about the distribution of IQ. It occurred to me that you could force the distribution to nearly whatever you liked simply by selecting certain questions. For example the distribution of cognitive ability would seem to be very different if the one and only question were to derive E= mc2 from first principles instead of 1 + 1 =? Yet, what you are really trying to do with an IQ test is to provide questions of increasing complexity that will pick up differences in ability levels. The two questions above really do not achieve this objective. With 1 + 1, presumably most would not need a calculator. One could then gradually escalate the complexity involved. Perhaps at the next graduation, one might ask what 1 +
    4 = and so on. The complexity level could then continue to escalate. At the top level, perhaps you could provide a 15 digit number and ask for the 8th root.
    When this construction of IQ test is used a normal distribution results.
    Starting at the low end, more and more people would be able to answer easy questions such as 7+8, until you reached a point where average people might be challenged for example 17*31. From this level forward, fewer and fewer would be able to provide the correct answer.

    One counterpoint that I do find troubling is that one of our teachers became quite distressed that the normal curve did not seem to apply in her class. She was called into the principal's office to explain this discrepancy. In her defence, she noted that the Asian students had scored so high that the entire distribution was non-normal. I do not feel comfortable with such a dogmatic stance: you really shouldn't be expected to provide the expected empirical answer when your observed answer is different. In a science course that would be simply fixing your labs. If enough people buy into the group think then the answer that you wind up with is simply a self-fulfilling prophecy.

    One aspect of the EA3 paper that has strangely wentunmentioned is that this GWAS has now brought this new EA research era to main Street. Let's face it, there are likely only a few global organizations that could meaningfully contribute to a 1.1 million person GWAS. It is now a quite small and intimate closet full of people doing this science.
    Now think of who might have the resources to do a 75 person study. This size was noted in the paper as being able to confirm the EA3 results. Basically, almost any organization could scrounge up the money necessary to do such a study. Probably even high schools would have the funds needed for their students to do this research.

    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  100. APilgrim says:
    @res
    Yes, because the IQ curve looks like that in reality. Do you think the IQ curve in the US is multimodal? If so, evidence? It is certainly not multimodal to the extreme of the strawman graphic you present.

    I agree that there are some distortions to normality (that was pretty much the point of my comment, if you read the whole thing, though I would have thought you would at least have noticed the very first sentence!) in the US IQ distribution due to different subpopulations. The major effects I see from that are:
    - Disproportionate representation for various groups at both extremes of the distribution.
    - Possible fat far right tail due to high IQ subpopulations. Which may be quite important if extremely high IQ is important for innovation (which I happen to think is the case).

    As far as I can tell, within +-2SD (roughly 95% of people) of the mean the normal distribution provides a pretty good model for the overall US IQ distribution.

    That you failed to understand the nuances of my full comment says much more about your reading comprehension than it does about my probability and statistics background.

    P.S. I take it you are staking out the "A small difference between the model and reality means the model is worthless" end of the false dichotomy.

    You take it wrong. Models serve constructive purposes, as do graphics. I routinely start with a simple model and/or engineering approximation. Then advance from heuristics to qualitative math & more advanced quantitative models. I may carry solutions to a closed form exact answer, which can be exactly modeled, and experimentally verified, if the project calls for it.

    American Caucasians, of European extraction, had a pronounced bimodal distribution, from the beginning of the IQ studies, until the late 1970s. At that point, most of these subjects, were banned by the Political Correct (PC) ‘thought-police’. IQ Tests and similar protocols such as SAT, ACT, MCAT, LSAT and the like were widely used to ‘weed-out’, the stupid white kids, from top academic programs. These tests clearly favored the bright kids among Jews & Asians. They also virtually eliminated Blacks from consideration. Low IQ Blacks have been shoe-horned into lots of academic programs, for which they are not qualified.

    Read More
    • Replies: @res

    American Caucasians, of European extraction, had a pronounced bimodal distribution, from the beginning of the IQ studies, until the late 1970s.
     
    Evidence?
    , @Factorize
    This is a slightly different normal curve with a different meaning, though it is strikingly normal.
    The bottom graphs labeled all include 2 million people at least in the top 5% measured over 35 years.
    Even when looking at men and women at the 5% ability level, there is normality between math and verbal tilt. It is as not obviously true to me that this normality would arise, though both the math and verbal should individually be normal. It is only when you look at strongly selected highly achieving groups do you see a bimodal distribution. High achieving men are almost exclusively quants (though there was considerable discussion about what high achieving actually meant- high total score, high math or high verbal)

    http://www.unz.com/jthompson/tilting-at-sex-differences/
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  101. utu says:
    @hyperbola

    I have read somewhere that 88% GWAS hits usually come from the noncoding part of DNA in studies of all kinds of different traits. So what is the biological mechanism how noncoding SPNs affect a trait?
     
    For humans the estimates seem to be that protein-coding regions encompass on the order of 1% of DNA. There are other regions that are used to produce various important RNAs such as tRNA, microRNA, etc. Still other regions seem to be at least transcribed although no known (direct) functions are known and they may not be conserved evolutionarily. Other regions may show epigenetic effects (DNA methylation), but without identification of a specific functional activity. This is a very complex subject - you might find this useful.

    My guess would be that many of the "junk" regions have to do with the spatial structure and packaging of DNA and hence on expression of different protein-coding regions.

    Defining functional DNA elements in the human genome
    https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4035993/

    As for 10 million SNPs, this study only measured 1 million people. Can we expect 7000 times more SNPs when we expand to 7 billion humans? We already know that many of the SNPs seem to be in DNA regions that are not under evolutionary selection (see paper).

    I am looking at the problem form the mathematical end. I am not really interested in mechanism what SNPs do or not. It appears that nobody really does know what exactly SPNs form the noncoding regions do but everybody agrees that they are important because they must be included in polygenic scores to give the scores predictive power.

    I do not know how phone numbers associated with different people made those people to move to particular geographic locations or was it the others way around. I do not care what is the direction of the causality vector. But having a sufficiently large sample of people, their phone numbers and their geographic locations I can discover the rule of this association. And if I can confirm the rule on an independent validation sample I will be able to decode (lat,long) from phone numbers.

    This is really the bottom line of IQ nature-nurture controversy. What is the maximal prediction power with any polygenic score? Is it close to heritability based on twin studies or not. How this or that SNP make somebody smarter or not is secondary to me.

    As for 10 million SNPs, this study only measured 1 million people. Can we expect 7000 times more SNPs when we expand to 7 billion humans?

    What are you talking about? Things do not scale like that.

    Read More
    • Replies: @j2
    utu writes:
    "What is the maximal prediction power with any polygenic score? Is it close to heritability based on twin studies or not."

    As I already stated two times, do not expect a polygenic score to show the whole predictive power of the SNPs that are included in the polygenic score. As a polygenic score is an additive concept, it is in fact motivated only as the first order approximation. Therefore is can rather well be used for comparisons where the scores differ only slightly. The calculated value of how much polygenic scores explain of the trait understate how much is genetically determined by the SNPs included in the polygenic score. A more correct mathematical method is to determine by other means, like from twin studies, how much is genetically determined, and to determine from GWAS what SNPs have any effect on the trait. Then the conclusion is that those SNPs that have effect contribute together for the whole genetic effect on the trait. This is the third time I write this, will not write a fourth time.
    , @hyperbola

    I am looking at the problem form the mathematical end. I am not really interested in mechanism what SNPs do or not. It appears that nobody really does know what exactly SPNs form the noncoding regions do but everybody agrees that they are important because they must be included in polygenic scores to give the scores predictive power.
     
    Not a wise attitude. There is already evidence that some regions of DNA are NOT under evolutionary selection, that is, SNPs in such regions may be completely random and insignificant for any trait. The very low levels of sampling currently available and the necessity of thousands of SNPs to get any "predictive" power (even as an average), suggest that much of this "predictive power" is simply noise.

    As for "scaling of SNPs", since for a million people we have only measured about 1 part per 7000 of human DNA (and that part is highly weighted towards certain populations), I doubt anyone knows how the scaling goes.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  102. @utu
    The frequencies f0,f1,f2, where f0+f1+f2=1 for any allele in any population that indicate frequency (probability) of occurrence of 0 allele, of 1 allele and of 2 allies, respectively are empirical values for a given sample. These frequencies suffice to determine the first moment of the distribution of the predicted trait by a polygenic score. But to determine the 2nd moment the frequencies alone are insufficient and a covariance matrix of SNPs included in the polygenic score must be used. The distribution is determined by a sample of N individuals. The polygenic score if only implicitly must be spelled out for each individual and in this score the frequencies do not play any role.

    If you want to connect genetically (directly caused) individual IQ, as a random variable, to (genetically caused) group IQ, then such a more careful calculation is needed to check the correctness of the result obtained from using only the means and frequencies of genes.

    This issue is dealt with by handwaving about units of aggregation, without offering hypotheses as to what may account for discrepancies between levels of aggregation. (The link is from the HBD reading list).

    I do not per se object to such aggregation, and it is unavoidable in studies of low frequency systems e.g. crime, whether one takes genetic, environmental or combined approaches.

    If one’s aim is to check whether group means can be predicted based on genes, without positing a hypothesis about individual scores, the approach is fine, as long as one does not pretend that one has demonstrated things that were not demonstrated.

    Read More
    • Replies: @utu

    If one’s aim is to check whether group means can be predicted based on genes, without positing a hypothesis about individual scores, the approach is fine, as long as one does not pretend that one has demonstrated things that were not demonstrated.
     
    Yes and no. I mean the approach is fine but there is but. Almost any group means can be "predicted" using genes because the set of group means usually have very few elements (races or countries or professions) and there is almost infinite number of combinations of SNPs, so a chance of finding one that correlates with group means is very high. This is what the defenders of David Piffer did not understand. His polygenic score of 9 or so SNPs correlated very well with IQ's of 20 or 30 countries. But one could assign to these countries virtually any set of random numbers instead of IQ's and a chance would be vey high that among 10 millions of SNPs there would be a subset that would produce a polygenic score correlating with these random numbers.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  103. res says:
    @APilgrim
    You take it wrong. Models serve constructive purposes, as do graphics. I routinely start with a simple model and/or engineering approximation. Then advance from heuristics to qualitative math & more advanced quantitative models. I may carry solutions to a closed form exact answer, which can be exactly modeled, and experimentally verified, if the project calls for it.

    American Caucasians, of European extraction, had a pronounced bimodal distribution, from the beginning of the IQ studies, until the late 1970s. At that point, most of these subjects, were banned by the Political Correct (PC) 'thought-police'. IQ Tests and similar protocols such as SAT, ACT, MCAT, LSAT and the like were widely used to 'weed-out', the stupid white kids, from top academic programs. These tests clearly favored the bright kids among Jews & Asians. They also virtually eliminated Blacks from consideration. Low IQ Blacks have been shoe-horned into lots of academic programs, for which they are not qualified.

    American Caucasians, of European extraction, had a pronounced bimodal distribution, from the beginning of the IQ studies, until the late 1970s.

    Evidence?

    Read More
    • Replies: @Tim too
    This is purely speculation, or hypothesizing on my part. In the event of a bimodal distribution, I would look into family culture effects, such as children of professional class parents (eg doctors, dentists, and other more highly educated parents) vs children of lower educated, non-technical classes, without a background of scholarship, somewhat loosely defined. Children of mechanics, for example might just perform similarly to children of engineers (because of the technical background, the search for technical solutions, even though there is a vastly larger math experience in engineering), I don't know. Just speculating.

    How does the parent think and problem solve? - leading to, how does the child think and perform in problem solving? A possible family culture effect. Children learn how to do things from their parents, as well as picking things up from peers, and media, etc.

    And so a test could be to sample by parent profession, education etc, (or by parent IQ) and examine child IQ distribution. I haven't read all the above comments to know if this has already been suggested.

    , @APilgrim
    Why not simultaneously disprove the anthropo(morphic/genic) 'climate-change' hoax?

    Why not correct the cosmology errors on the back of a napkin, while I am at it? ...

    I studied these matters ... B4 the PC Police 'burned the books'. And I recall the answers.

    I am neither your schoolmarm, nor your minion.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  104. Tim too says:
    @res

    American Caucasians, of European extraction, had a pronounced bimodal distribution, from the beginning of the IQ studies, until the late 1970s.
     
    Evidence?

    This is purely speculation, or hypothesizing on my part. In the event of a bimodal distribution, I would look into family culture effects, such as children of professional class parents (eg doctors, dentists, and other more highly educated parents) vs children of lower educated, non-technical classes, without a background of scholarship, somewhat loosely defined. Children of mechanics, for example might just perform similarly to children of engineers (because of the technical background, the search for technical solutions, even though there is a vastly larger math experience in engineering), I don’t know. Just speculating.

    How does the parent think and problem solve? – leading to, how does the child think and perform in problem solving? A possible family culture effect. Children learn how to do things from their parents, as well as picking things up from peers, and media, etc.

    And so a test could be to sample by parent profession, education etc, (or by parent IQ) and examine child IQ distribution. I haven’t read all the above comments to know if this has already been suggested.

    Read More
    • Replies: @Tim too
    and to add: sample by parental pressure on children to excel!
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  105. Factorize says:
    @APilgrim
    You take it wrong. Models serve constructive purposes, as do graphics. I routinely start with a simple model and/or engineering approximation. Then advance from heuristics to qualitative math & more advanced quantitative models. I may carry solutions to a closed form exact answer, which can be exactly modeled, and experimentally verified, if the project calls for it.

    American Caucasians, of European extraction, had a pronounced bimodal distribution, from the beginning of the IQ studies, until the late 1970s. At that point, most of these subjects, were banned by the Political Correct (PC) 'thought-police'. IQ Tests and similar protocols such as SAT, ACT, MCAT, LSAT and the like were widely used to 'weed-out', the stupid white kids, from top academic programs. These tests clearly favored the bright kids among Jews & Asians. They also virtually eliminated Blacks from consideration. Low IQ Blacks have been shoe-horned into lots of academic programs, for which they are not qualified.

    This is a slightly different normal curve with a different meaning, though it is strikingly normal.
    The bottom graphs labeled all include 2 million people at least in the top 5% measured over 35 years.
    Even when looking at men and women at the 5% ability level, there is normality between math and verbal tilt. It is as not obviously true to me that this normality would arise, though both the math and verbal should individually be normal. It is only when you look at strongly selected highly achieving groups do you see a bimodal distribution. High achieving men are almost exclusively quants (though there was considerable discussion about what high achieving actually meant- high total score, high math or high verbal)

    http://www.unz.com/jthompson/tilting-at-sex-differences/

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  106. Tim too says:
    @Tim too
    This is purely speculation, or hypothesizing on my part. In the event of a bimodal distribution, I would look into family culture effects, such as children of professional class parents (eg doctors, dentists, and other more highly educated parents) vs children of lower educated, non-technical classes, without a background of scholarship, somewhat loosely defined. Children of mechanics, for example might just perform similarly to children of engineers (because of the technical background, the search for technical solutions, even though there is a vastly larger math experience in engineering), I don't know. Just speculating.

    How does the parent think and problem solve? - leading to, how does the child think and perform in problem solving? A possible family culture effect. Children learn how to do things from their parents, as well as picking things up from peers, and media, etc.

    And so a test could be to sample by parent profession, education etc, (or by parent IQ) and examine child IQ distribution. I haven't read all the above comments to know if this has already been suggested.

    and to add: sample by parental pressure on children to excel!

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  107. utu says:
    @Johan Meyer
    If you want to connect genetically (directly caused) individual IQ, as a random variable, to (genetically caused) group IQ, then such a more careful calculation is needed to check the correctness of the result obtained from using only the means and frequencies of genes.

    This issue is dealt with by handwaving about units of aggregation, without offering hypotheses as to what may account for discrepancies between levels of aggregation. (The link is from the HBD reading list).

    I do not per se object to such aggregation, and it is unavoidable in studies of low frequency systems e.g. crime, whether one takes genetic, environmental or combined approaches.

    If one's aim is to check whether group means can be predicted based on genes, without positing a hypothesis about individual scores, the approach is fine, as long as one does not pretend that one has demonstrated things that were not demonstrated.

    If one’s aim is to check whether group means can be predicted based on genes, without positing a hypothesis about individual scores, the approach is fine, as long as one does not pretend that one has demonstrated things that were not demonstrated.

    Yes and no. I mean the approach is fine but there is but. Almost any group means can be “predicted” using genes because the set of group means usually have very few elements (races or countries or professions) and there is almost infinite number of combinations of SNPs, so a chance of finding one that correlates with group means is very high. This is what the defenders of David Piffer did not understand. His polygenic score of 9 or so SNPs correlated very well with IQ’s of 20 or 30 countries. But one could assign to these countries virtually any set of random numbers instead of IQ’s and a chance would be vey high that among 10 millions of SNPs there would be a subset that would produce a polygenic score correlating with these random numbers.

    Read More
    • Replies: @Johan Meyer
    In other words, more than one set of SNPs could be used to predict group IQs.

    Then a very simple approach is available.

    1. Find more countries, or subpopulations (e.g. over time and generations).
    2. Remove the prediction sets that fail to account for the enlarged set of IQ data.
    3. If viable sets remain, go to 1.

    If, as I have suggested, the gene connection is due to variable uptake resulting in an effective variable dose response to external dose of environment, then step 1 should over repeated cycles of observing new generations empty the prediction set. Ditto if the result is simply sporadic.

    Alternatively, the length of the remaining viable sets may continue to grow. And in this regard, it may be useful to check the lengths of sets, over time (as new data is added) to maintain correlations of 0.7, 0.8, and 0.9. If these continue to grow as data is collected, then we may be sure that the process is not (perhaps no longer) latching on to a real phenomenon.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  108. APilgrim says:
    @res

    American Caucasians, of European extraction, had a pronounced bimodal distribution, from the beginning of the IQ studies, until the late 1970s.
     
    Evidence?

    Why not simultaneously disprove the anthropo(morphic/genic) ‘climate-change’ hoax?

    Why not correct the cosmology errors on the back of a napkin, while I am at it? …

    I studied these matters … B4 the PC Police ‘burned the books’. And I recall the answers.

    I am neither your schoolmarm, nor your minion.

    Read More
    • LOL: res
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  109. j2 says:
    @Factorize
    j2, the normal distribution for IQ arose from empirical observations. Pschometricians looked out in the community and saw that most people were of average intelligence and some were of greater and some were of lower intelligence. A normal distribution results when you add up many random variables as with IQ. It is the logical outcome for the genetic architecture of intelligence that has been found.

    The idea of a multimodal distribution is not supported by the evidence. Recent research has found that different socioeconomic groups are not that exceptionally different. Reshuffling of the SNPs every generation tends to pull everyone back to average.

    Life would be structured much differently if this were not true. If there were a multimodal IQ distribution it would have been much more necessary to have separate schools for those of differing ability levels. This was by and large not found to be necessary.

    Look, Factorize, safely assume I know all about basics of statistics or any basic math and have proven all those theorems in exercises ages ago. We know that we get the normal distribution under weak conditions, but if you have an IQ test with a set of questions and the distribution is not normal, you assume that, sorry, it has to be normal as it is normal under weak conditions, so my test set is just poorly chosen. Like, make an IQ test set with half trivially simple questions and half unbelievably difficult questions, you get the distribution that almost all got only the trivially simple questions, but one guy (who stole the test questions in advance) got everything correct. This is empirical, and you conclude that empirical is not correct, the test set was poorly chosen, if correctly chosen it must give normal. And this is not because of engineers, they do not know anything for sure, it is mathematicians, they have all the proofs.

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  110. j2 says:
    @utu
    I am looking at the problem form the mathematical end. I am not really interested in mechanism what SNPs do or not. It appears that nobody really does know what exactly SPNs form the noncoding regions do but everybody agrees that they are important because they must be included in polygenic scores to give the scores predictive power.

    I do not know how phone numbers associated with different people made those people to move to particular geographic locations or was it the others way around. I do not care what is the direction of the causality vector. But having a sufficiently large sample of people, their phone numbers and their geographic locations I can discover the rule of this association. And if I can confirm the rule on an independent validation sample I will be able to decode (lat,long) from phone numbers.

    This is really the bottom line of IQ nature-nurture controversy. What is the maximal prediction power with any polygenic score? Is it close to heritability based on twin studies or not. How this or that SNP make somebody smarter or not is secondary to me.

    As for 10 million SNPs, this study only measured 1 million people. Can we expect 7000 times more SNPs when we expand to 7 billion humans?
     
    What are you talking about? Things do not scale like that.

    utu writes:
    “What is the maximal prediction power with any polygenic score? Is it close to heritability based on twin studies or not.”

    As I already stated two times, do not expect a polygenic score to show the whole predictive power of the SNPs that are included in the polygenic score. As a polygenic score is an additive concept, it is in fact motivated only as the first order approximation. Therefore is can rather well be used for comparisons where the scores differ only slightly. The calculated value of how much polygenic scores explain of the trait understate how much is genetically determined by the SNPs included in the polygenic score. A more correct mathematical method is to determine by other means, like from twin studies, how much is genetically determined, and to determine from GWAS what SNPs have any effect on the trait. Then the conclusion is that those SNPs that have effect contribute together for the whole genetic effect on the trait. This is the third time I write this, will not write a fourth time.

    Read More
    • Replies: @utu
    You would not ave to write things 2 or 3 times if you knew that that it was unnecessary writing even the first time.

    Comment #17
    The problem is complex. When using linear polygenic score it becomes much simpler mathematically. But there are simplifications which I am not sure that are justified. Presumably they might be justified by GWAS. I do not know. For example a linear polygenic score implies that the additive effect from 0 allele to 1 allele change is the same as from 1 allele two 2 alleles. Then it implies that different SNPs are ignorant of each other while it might be possible that the effect of having SNP1 and SNP2 is not a sum of effects of the two SNPs together. To discover it one would hav to either do non-linear polygenic score or even better to an equivalent of logistic regression where SNPs cannot be mathematized to numerical values.
     
    So obviously when I used the term polygenic score in arguing a more general point its linearity was not implied.

    This is the third time I write this, will not write a fourth time.
     
    What was it you were trying to explain?
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  111. Okechukwu says:
    @Anon
    > African immigrants are more educated than most — including people born in U.S.

    Lynn and Vanhanen showed the IQ in the native countries. You tried to counter that with specially selected elite African immigrants with that?? You do not seem to understand selection bias.

    You are a pathetic loser.

    Lynn and Vanhanen showed the IQ in the native countries.

    Lynn and Vanhanen fabricated IQ data in the native countries. Unless you believe that a group of retarded children in Spain provide an accurate proxy for the people of Equatorial Guinea (among many other absurdities), then even a moron like you would have to admit that the Lynn/Vanhenen data is fraudulent.

    Actually, those pseudoscientific charlatans were counting precisely on ignorant people like you to receive their “research.” But unfortunately for them, some real scientists got ahold of their data and proceeded to rip them new assholes. Beyond that, even laymen understand that Africans who speak multiple languages, function normally and perform every endeavor no matter how intellectually rigorous, CANNOT be functionally retarded. As a consequence, the only people that place any faith in the fraudulent Lynn/Vanhanen material are racist Internet white power trolls. For the rest of the global population, it’s completely worthless and irrelevant. That’s why you’ll never see it quoted, cited or referenced by any credible source anywhere in the world. You see, outside of these echo-chambers and in the real world, we do this thing called due diligence. It’s the thing that consigns all of these theories to the garbage heap.

    You tried to counter that with specially selected elite African immigrants with that?? You do not seem to understand selection bias.

    There is no selection bias in African immigration. Immigrants represent a cross-section of African society and not some elite. You know, it’s really hard to get an elite via the visa lottery and chain migration. People who sponsor their relatives don’t care how smart or dumb they are. Besides, basic common sense should tell you that you cannot get an elite capable of grossly outperforming whites from a source population of 55 IQ retards.

    Read More
    • Replies: @res

    There is no selection bias in African immigration. Immigrants represent a cross-section of African society and not some elite.
     
    LOL! Here we go with this one again. Let's take a look at the IAB Brain Drain data for immigration once more. I have shown you this enough times it is clear you are just a liar.

    https://www.iab.de/en/daten/iab-brain-drain-data.aspx

    The 2010 emigration rates (men and women combined) from Nigeria by educational level were:
    Total   | Low    | Medium | High
    0.61% | 0.11% | 0.56%     | 12.04%

    Nope, not selective at all. Just a 100:1 ratio of emigration between the high and low groups.

    It is fun to note that the credentials brought by African immigrants to the US which you trumpet so loudly as proof of excellence is exactly what disproves your statement about immigration selectivity.

    Besides, basic common sense should tell you that you cannot get an elite capable of grossly outperforming whites from a source population of 55 IQ retards.
     
    So it's 55 IQ now? You should not speak so ill of the country of your ancestral origin. Even your hated Lynn did not give an estimate that low. Although his 69 for Nigeria does seem rather pessimistic: https://pdfs.semanticscholar.org/48d7/843f6ce714a684a93530a0c8b7da65d185db.pdf
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  112. @utu

    If one’s aim is to check whether group means can be predicted based on genes, without positing a hypothesis about individual scores, the approach is fine, as long as one does not pretend that one has demonstrated things that were not demonstrated.
     
    Yes and no. I mean the approach is fine but there is but. Almost any group means can be "predicted" using genes because the set of group means usually have very few elements (races or countries or professions) and there is almost infinite number of combinations of SNPs, so a chance of finding one that correlates with group means is very high. This is what the defenders of David Piffer did not understand. His polygenic score of 9 or so SNPs correlated very well with IQ's of 20 or 30 countries. But one could assign to these countries virtually any set of random numbers instead of IQ's and a chance would be vey high that among 10 millions of SNPs there would be a subset that would produce a polygenic score correlating with these random numbers.

    In other words, more than one set of SNPs could be used to predict group IQs.

    Then a very simple approach is available.

    1. Find more countries, or subpopulations (e.g. over time and generations).
    2. Remove the prediction sets that fail to account for the enlarged set of IQ data.
    3. If viable sets remain, go to 1.

    If, as I have suggested, the gene connection is due to variable uptake resulting in an effective variable dose response to external dose of environment, then step 1 should over repeated cycles of observing new generations empty the prediction set. Ditto if the result is simply sporadic.

    Alternatively, the length of the remaining viable sets may continue to grow. And in this regard, it may be useful to check the lengths of sets, over time (as new data is added) to maintain correlations of 0.7, 0.8, and 0.9. If these continue to grow as data is collected, then we may be sure that the process is not (perhaps no longer) latching on to a real phenomenon.

    Read More
    • Replies: @Johan Meyer
    The sets used to validate (avoid unjustified prediction sequence growth) as more data is included, i.e. to achieve 0.7, 0.8 and 0.9 correlations, should be subsets of the sequences used in the publications (papers) in which they appear. That should be adequate to avoid specious sequences.
    , @utu

    Then a very simple approach is available.
     
    Do you realize how big the number 2^10,000,000 is? This is the number of all possible subsets from 10,000,000 SNPs. There might be SNPs that will never be found by any GWAS when searched for individually because their correlations with the trait will be zero regardless of the sample size but the same SNPs as a group might be effect on a trait. I am not saying this is so or must be so but there is a mathematical possibility.

    You were right when you wrote that from the correlation established on averages of subsets we can't infer correlation on individuals.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  113. res says:
    @Okechukwu

    Lynn and Vanhanen showed the IQ in the native countries.
     
    Lynn and Vanhanen fabricated IQ data in the native countries. Unless you believe that a group of retarded children in Spain provide an accurate proxy for the people of Equatorial Guinea (among many other absurdities), then even a moron like you would have to admit that the Lynn/Vanhenen data is fraudulent.

    Actually, those pseudoscientific charlatans were counting precisely on ignorant people like you to receive their "research." But unfortunately for them, some real scientists got ahold of their data and proceeded to rip them new assholes. Beyond that, even laymen understand that Africans who speak multiple languages, function normally and perform every endeavor no matter how intellectually rigorous, CANNOT be functionally retarded. As a consequence, the only people that place any faith in the fraudulent Lynn/Vanhanen material are racist Internet white power trolls. For the rest of the global population, it's completely worthless and irrelevant. That's why you'll never see it quoted, cited or referenced by any credible source anywhere in the world. You see, outside of these echo-chambers and in the real world, we do this thing called due diligence. It's the thing that consigns all of these theories to the garbage heap.


    You tried to counter that with specially selected elite African immigrants with that?? You do not seem to understand selection bias.
     
    There is no selection bias in African immigration. Immigrants represent a cross-section of African society and not some elite. You know, it's really hard to get an elite via the visa lottery and chain migration. People who sponsor their relatives don't care how smart or dumb they are. Besides, basic common sense should tell you that you cannot get an elite capable of grossly outperforming whites from a source population of 55 IQ retards.

    There is no selection bias in African immigration. Immigrants represent a cross-section of African society and not some elite.

    LOL! Here we go with this one again. Let’s take a look at the IAB Brain Drain data for immigration once more. I have shown you this enough times it is clear you are just a liar.

    https://www.iab.de/en/daten/iab-brain-drain-data.aspx

    The 2010 emigration rates (men and women combined) from Nigeria by educational level were:
    Total   | Low    | Medium | High
    0.61% | 0.11% | 0.56%     | 12.04%

    Nope, not selective at all. Just a 100:1 ratio of emigration between the high and low groups.

    It is fun to note that the credentials brought by African immigrants to the US which you trumpet so loudly as proof of excellence is exactly what disproves your statement about immigration selectivity.

    Besides, basic common sense should tell you that you cannot get an elite capable of grossly outperforming whites from a source population of 55 IQ retards.

    So it’s 55 IQ now? You should not speak so ill of the country of your ancestral origin. Even your hated Lynn did not give an estimate that low. Although his 69 for Nigeria does seem rather pessimistic: https://pdfs.semanticscholar.org/48d7/843f6ce714a684a93530a0c8b7da65d185db.pdf

    Read More
    • Replies: @Okechukwu

    LOL! Here we go with this one again. Let’s take a look at the IAB Brain Drain data for immigration once more. I have shown you this enough times it is clear you are just a liar.
     
    Brain drain pertains to educated professionals, not necessarily to the most intelligent. But even as regards educated professionals, brain drain is largely a myth. The supply of educated Nigerians is not finite. They have a robust post-secondary system out of which most graduates stay in Nigeria. Also, many return to Nigeria after their academic work is completed in the west. Particularly with Igbos, the idea of being buried in a foreign land is anathema. If they don't make it while alive, their bodies are certainly sent back if they die here.

    The smartest Nigerians are actually in Nigeria. Immigrating to the United States is extremely challenging and most don't have the means, the wherewithal or the family connections to do so. It only seems like the Nigerians in America are elite because America compares very favorably to Nigeria in terms of the ability to achieve by dint of hard work and perseverance. Having been exposed to a hardscrabble, difficult experience in Nigeria, America becomes a relatively easy downhill ride.


    It is fun to note that the credentials brought by African immigrants to the US which you trumpet so loudly as proof of excellence is exactly what disproves your statement about immigration selectivity.
     
    Those credentials are acquired in the United States, otherwise they would be worthless. All references to the educational attainment of Africans in America refer exclusively to degrees acquired at American institutions. This is the point I'm trying to get across to you. Most Nigerian immigrants don't arrive as elites. They become elites in the United States.

    So it’s 55 IQ now? You should not speak so ill of the country of your ancestral origin. Even your hated Lynn did not give an estimate that low. Although his 69 for Nigeria does seem rather pessimistic:
     
    But on the map whose veracity you seem to be vouching for, some African countries are assigned IQ's that low. And even 69 would meet the standard for severe mental disability. That would mean that all Nigerian Olympic athletes would be shifted to the Special Olympics. It would also mean that Nigerians probably couldn't operate an automobile or brush their own teeth. So racist pseudoscientists like Lynn were hoisted by their own petard. Believing that their audience wouldn't extend beyond white nationalist simpletons, they went much too far in their mission to establish the intellectual inferiority of Africans. In fact due to negative peer reviews, Lynn has yet again revised his IQ estimate for Nigeria, this time to 84. He basically just pulls numbers out of his ass.

    https://iq-research.info/en/average-iq-by-country

    Furthermore, according to Lynn, Sierra Leone with an IQ of 91 scores higher than a host of European, Latin American, Middle Eastern and Asian countries. Kind of runs counter to the HBD narrative, doesn't it?

    Here's the thing, anyone that tells you they have credible IQ scores for any African country is lying. No valid, properly controlled, large sample size survey has ever been done. In most cases the scores were simply extrapolated against a backdrop of the researcher's own deeply ingrained biases. At other times they would walk into a village and test a few illiterate, malnourished kids and on that basis arrive at an average for the country as a whole, making sure of course to weed out all high scores as anomalous. With this kind of methodology I could get an average IQ of 50 for white Americans. I could simply test attendees at a Trump rally or I could test the inhabitants of backwoods Appalachia, making sure to throw out any scores above 50.

    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  114. APilgrim says:

    Until the 1980s, UT Austin automatically accepted freshmen admits on the baseis of: Top 10% of SAT/ACT or High School Class Rank. Some legacies, big donor, and star athletes were ‘worked-in’. Graduates of low performing High Schools were excluded from Automatic Entry. This system cumulatively resulted in few Black admissions. A couple of quota programs were focused on Black, Hispanic & AmerIND admissions. Extensive data was mandated. https://provost.utexas.edu/enrollment-management/admissions-research The Quota-Kids could not keep up; court decisions were mixed, and Texas voters went on the warpath. The last gasp Quota Plan, (AKA Diversity-Plan, which included deviates) was unlawfully adopted by the entire Texas higher education system.

    The Texas Legislature fired and/or ordered prosecution of all the university presidents & chancellors. The Lege subsequently mandated that 75% of In-State Freshman Admits would be entirely based upon Texas High School Class Rank. For the 2018-2019 Academic year, that is the Top 7% of class rank, and for the 2019-2010 Academic year, a Top 6% class ranking is required. ‘Diversity’ is legally defined as towns (geography) in Texas. There is a holistic process available for the other 25% of In-State admissions. http://www.prepscholar.com/sat/s/colleges/UT-Austin-admission-requirements. This holistic allows athletes, savants, big-donors, & minorities to gain admissions.

    The University of Texas (Top-7%) Plan has survived two (2) round-trips to SCOTUS.

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter Display All Comments
  115. @Johan Meyer
    In other words, more than one set of SNPs could be used to predict group IQs.

    Then a very simple approach is available.

    1. Find more countries, or subpopulations (e.g. over time and generations).
    2. Remove the prediction sets that fail to account for the enlarged set of IQ data.
    3. If viable sets remain, go to 1.

    If, as I have suggested, the gene connection is due to variable uptake resulting in an effective variable dose response to external dose of environment, then step 1 should over repeated cycles of observing new generations empty the prediction set. Ditto if the result is simply sporadic.

    Alternatively, the length of the remaining viable sets may continue to grow. And in this regard, it may be useful to check the lengths of sets, over time (as new data is added) to maintain correlations of 0.7, 0.8, and 0.9. If these continue to grow as data is collected, then we may be sure that the process is not (perhaps no longer) latching on to a real phenomenon.

    The sets used to validate (avoid unjustified prediction sequence growth) as more data is included, i.e. to achieve 0.7, 0.8 and 0.9 correlations, should be subsets of the sequences used in the publications (papers) in which they appear. That should be adequate to avoid specious sequences.

    Read More
    • Replies: @res

    should be subsets of the sequences used in the publications (papers) in which they appear.
     
    Better if they are complete sets or subsets created using a well defined a priori criteria. Otherwise you run into issues with multiple hypothesis testing.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  116. utu says:
    @j2
    utu writes:
    "What is the maximal prediction power with any polygenic score? Is it close to heritability based on twin studies or not."

    As I already stated two times, do not expect a polygenic score to show the whole predictive power of the SNPs that are included in the polygenic score. As a polygenic score is an additive concept, it is in fact motivated only as the first order approximation. Therefore is can rather well be used for comparisons where the scores differ only slightly. The calculated value of how much polygenic scores explain of the trait understate how much is genetically determined by the SNPs included in the polygenic score. A more correct mathematical method is to determine by other means, like from twin studies, how much is genetically determined, and to determine from GWAS what SNPs have any effect on the trait. Then the conclusion is that those SNPs that have effect contribute together for the whole genetic effect on the trait. This is the third time I write this, will not write a fourth time.

    You would not ave to write things 2 or 3 times if you knew that that it was unnecessary writing even the first time.

    Comment #17
    The problem is complex. When using linear polygenic score it becomes much simpler mathematically. But there are simplifications which I am not sure that are justified. Presumably they might be justified by GWAS. I do not know. For example a linear polygenic score implies that the additive effect from 0 allele to 1 allele change is the same as from 1 allele two 2 alleles. Then it implies that different SNPs are ignorant of each other while it might be possible that the effect of having SNP1 and SNP2 is not a sum of effects of the two SNPs together. To discover it one would hav to either do non-linear polygenic score or even better to an equivalent of logistic regression where SNPs cannot be mathematized to numerical values.

    So obviously when I used the term polygenic score in arguing a more general point its linearity was not implied.

    This is the third time I write this, will not write a fourth time.

    What was it you were trying to explain?

    Read More
    • Replies: @j2
    "What was it you were trying to explain?"

    Point taken, so I try to write something more reasonable to better address your comment. To get the polygenic score method more predictive you have to find SNP combinations that effect IQ. Separate from the samples high and low performance subsets and try to identify with 2 SNP or 3 SNP combinations appear more often. Then continue this way as far as it goes, maybe to 4 SNP combinations.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  117. res says:
    @Johan Meyer
    The sets used to validate (avoid unjustified prediction sequence growth) as more data is included, i.e. to achieve 0.7, 0.8 and 0.9 correlations, should be subsets of the sequences used in the publications (papers) in which they appear. That should be adequate to avoid specious sequences.

    should be subsets of the sequences used in the publications (papers) in which they appear.

    Better if they are complete sets or subsets created using a well defined a priori criteria. Otherwise you run into issues with multiple hypothesis testing.

    Read More
    • Replies: @Johan Meyer
    The criteria are achieving the correlations (R or R^2, as previously achieved) of 0.7, 0.8 and 0.9.

    One cannot decrease R^2 by adding explanatory variables. Thus, the question is whether the added explanatory variables used to fit a bigger data set are because of greater insight offered by the additional data, or whether it is simply finding a random pattern that happens to match tolerably well.

    If understanding is truely advancing, as suggested by increasing R^2 over time, then one should expect that lesser R^2s can be achieved by a small set, even as the data grows with time.

    Thus the challenge is to find the smallest set (preferably with the same coefficients/betas as the study being published) that will yield R^2s of at least 0.7, 0.8, and 0.9.

    If the sizes of these minimal R^2 achieving sets grow as data is added, it raises the concern that one is achieving better R^2s by inflating the explanatory variables from a random set, rather than achieving understanding.

    My suspicion is that polygenic scores are real, and that the set sizes will not grow for sets taken prior to lead clean up, as I suspect that it is a polygenic score for uptake of lead. If the sizes fail to grow even after lead clean up, it would tend to point to direct genetic contribution to IQ.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  118. @res

    should be subsets of the sequences used in the publications (papers) in which they appear.
     
    Better if they are complete sets or subsets created using a well defined a priori criteria. Otherwise you run into issues with multiple hypothesis testing.

    The criteria are achieving the correlations (R or R^2, as previously achieved) of 0.7, 0.8 and 0.9.

    One cannot decrease R^2 by adding explanatory variables. Thus, the question is whether the added explanatory variables used to fit a bigger data set are because of greater insight offered by the additional data, or whether it is simply finding a random pattern that happens to match tolerably well.

    If understanding is truely advancing, as suggested by increasing R^2 over time, then one should expect that lesser R^2s can be achieved by a small set, even as the data grows with time.

    Thus the challenge is to find the smallest set (preferably with the same coefficients/betas as the study being published) that will yield R^2s of at least 0.7, 0.8, and 0.9.

    If the sizes of these minimal R^2 achieving sets grow as data is added, it raises the concern that one is achieving better R^2s by inflating the explanatory variables from a random set, rather than achieving understanding.

    My suspicion is that polygenic scores are real, and that the set sizes will not grow for sets taken prior to lead clean up, as I suspect that it is a polygenic score for uptake of lead. If the sizes fail to grow even after lead clean up, it would tend to point to direct genetic contribution to IQ.

    Read More
    • Replies: @Johan Meyer
    A further criterion for such sets would be that if they differ (either in betas for the same SNPs, or in choice of SNPs), that they must also produce the desired minimum correlations for smaller data sets of previous publications.
    , @res
    Simple question. Do you understand what "multiple hypothesis testing" means?

    If so, do you see how it applies to choosing subsets of SNPs for a PGS?

    If not, this might help: https://en.wikipedia.org/wiki/Multiple_comparisons_problem
    , @Chainsaw1
    Now learn your stats 102.

    https://en.wikipedia.org/wiki/Coefficient_of_determination#Adjusted_R2

    "If the sizes of these minimal R^2 achieving sets grow as data is added, it raises the concern that one is achieving better R^2s by inflating the explanatory variables from a random set, rather than achieving understanding. Unlike R2, the adjusted R2 increases only when the increase in R2 (due to the inclusion of a new explanatory variable) is more than one would expect to see by chance."

    Now for the stats 103.

    https://en.wikipedia.org/wiki/Akaike_information_criterion

    "AIC rewards goodness of fit (as assessed by the likelihood function), but it also includes a penalty that is an increasing function of the number of estimated parameters. The penalty discourages overfitting, because increasing the number of parameters in the model almost always improves the goodness of the fit."
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  119. utu says:
    @Johan Meyer
    In other words, more than one set of SNPs could be used to predict group IQs.

    Then a very simple approach is available.

    1. Find more countries, or subpopulations (e.g. over time and generations).
    2. Remove the prediction sets that fail to account for the enlarged set of IQ data.
    3. If viable sets remain, go to 1.

    If, as I have suggested, the gene connection is due to variable uptake resulting in an effective variable dose response to external dose of environment, then step 1 should over repeated cycles of observing new generations empty the prediction set. Ditto if the result is simply sporadic.

    Alternatively, the length of the remaining viable sets may continue to grow. And in this regard, it may be useful to check the lengths of sets, over time (as new data is added) to maintain correlations of 0.7, 0.8, and 0.9. If these continue to grow as data is collected, then we may be sure that the process is not (perhaps no longer) latching on to a real phenomenon.

    Then a very simple approach is available.

    Do you realize how big the number 2^10,000,000 is? This is the number of all possible subsets from 10,000,000 SNPs. There might be SNPs that will never be found by any GWAS when searched for individually because their correlations with the trait will be zero regardless of the sample size but the same SNPs as a group might be effect on a trait. I am not saying this is so or must be so but there is a mathematical possibility.

    You were right when you wrote that from the correlation established on averages of subsets we can’t infer correlation on individuals.

    Read More
    • Replies: @Johan Meyer
    A simple hypothesis may explain the discrepancy between group (aggregate) and individual effects, to wit, genetically variable sensitivity to environment, with shared aggregate environment being negatively correlated with the genes varying sensitivity to environment. Thus more sensitive individuals may be more concentrated in places where the pathological environment is less of an issue.

    Other hypotheses could also explain such scenarios, e.g. additive hypotheses with missing variables, or failure to include individual-level second moments (covariance matrices).

    The problem is precisely the lack of adequate specification of hypotheses.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  120. @Johan Meyer
    The criteria are achieving the correlations (R or R^2, as previously achieved) of 0.7, 0.8 and 0.9.

    One cannot decrease R^2 by adding explanatory variables. Thus, the question is whether the added explanatory variables used to fit a bigger data set are because of greater insight offered by the additional data, or whether it is simply finding a random pattern that happens to match tolerably well.

    If understanding is truely advancing, as suggested by increasing R^2 over time, then one should expect that lesser R^2s can be achieved by a small set, even as the data grows with time.

    Thus the challenge is to find the smallest set (preferably with the same coefficients/betas as the study being published) that will yield R^2s of at least 0.7, 0.8, and 0.9.

    If the sizes of these minimal R^2 achieving sets grow as data is added, it raises the concern that one is achieving better R^2s by inflating the explanatory variables from a random set, rather than achieving understanding.

    My suspicion is that polygenic scores are real, and that the set sizes will not grow for sets taken prior to lead clean up, as I suspect that it is a polygenic score for uptake of lead. If the sizes fail to grow even after lead clean up, it would tend to point to direct genetic contribution to IQ.

    A further criterion for such sets would be that if they differ (either in betas for the same SNPs, or in choice of SNPs), that they must also produce the desired minimum correlations for smaller data sets of previous publications.

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  121. @utu

    Then a very simple approach is available.
     
    Do you realize how big the number 2^10,000,000 is? This is the number of all possible subsets from 10,000,000 SNPs. There might be SNPs that will never be found by any GWAS when searched for individually because their correlations with the trait will be zero regardless of the sample size but the same SNPs as a group might be effect on a trait. I am not saying this is so or must be so but there is a mathematical possibility.

    You were right when you wrote that from the correlation established on averages of subsets we can't infer correlation on individuals.

    A simple hypothesis may explain the discrepancy between group (aggregate) and individual effects, to wit, genetically variable sensitivity to environment, with shared aggregate environment being negatively correlated with the genes varying sensitivity to environment. Thus more sensitive individuals may be more concentrated in places where the pathological environment is less of an issue.

    Other hypotheses could also explain such scenarios, e.g. additive hypotheses with missing variables, or failure to include individual-level second moments (covariance matrices).

    The problem is precisely the lack of adequate specification of hypotheses.

    Read More
    • Replies: @utu

    A simple hypothesis may explain
     
    You seem to overuse the word "simple". Nothing is really simple in this problem. The mathematical problem is insanely undetermined. There are 2^10,000,000 of potential solutions. Steven Hsu reduced the size of the problem by looking for solutions within a 50,000 only subset of SNPs where each SNP in the subset had a detectable correlation with the trait. Then he applied L1-fit (Lasso method) to find the smallest subset among them that maximized the explained variance. He ended up with 10,000 SNPs in his linear polygenic score. But he did not prove this was the only solution. It is very likely there are solutions among 10,000,000-50k subset that he did not use not mentioning the nonlinear polygenic score solutions that perhaps may require less than 10,000 SNPs to achieve a fit of the same quality.

    Returning to the issue of averages of subsets. If you have two variables Y an X (long lists of numbers) they may not correlate with each other but you may subdivide them into subsets X_i, Y_i, where i=0,1,2,... and the averages and of the subsets may correlate with each other. This fact tells us nothing about the correlation or lack of it between X and Y.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  122. res says:
    @Johan Meyer
    The criteria are achieving the correlations (R or R^2, as previously achieved) of 0.7, 0.8 and 0.9.

    One cannot decrease R^2 by adding explanatory variables. Thus, the question is whether the added explanatory variables used to fit a bigger data set are because of greater insight offered by the additional data, or whether it is simply finding a random pattern that happens to match tolerably well.

    If understanding is truely advancing, as suggested by increasing R^2 over time, then one should expect that lesser R^2s can be achieved by a small set, even as the data grows with time.

    Thus the challenge is to find the smallest set (preferably with the same coefficients/betas as the study being published) that will yield R^2s of at least 0.7, 0.8, and 0.9.

    If the sizes of these minimal R^2 achieving sets grow as data is added, it raises the concern that one is achieving better R^2s by inflating the explanatory variables from a random set, rather than achieving understanding.

    My suspicion is that polygenic scores are real, and that the set sizes will not grow for sets taken prior to lead clean up, as I suspect that it is a polygenic score for uptake of lead. If the sizes fail to grow even after lead clean up, it would tend to point to direct genetic contribution to IQ.

    Simple question. Do you understand what “multiple hypothesis testing” means?

    If so, do you see how it applies to choosing subsets of SNPs for a PGS?

    If not, this might help: https://en.wikipedia.org/wiki/Multiple_comparisons_problem

    Read More
    • Replies: @Johan Meyer
    I read the article. You choose a number of hypotheses (SNPs being significant for IQ), and find an estimate of the number of erroneous "discoveries" (SNPs that appear to contribute but do not, based on randomness), and erroneous rejections (SNPs that appear to be irrelevant to IQ but in fact contribute, again due to randomness), both under the hypothesis of a per-SNP test error rate (alpha).

    That begs the question of the basis for estimation of such an error rate. A further hypothesis (perhaps constrained by real data) is needed. Do you wish to posit an explicit hypothesis, perhaps with falsification criteria?
    , @Johan Meyer
    I should also add the following. Assume a p score of 10^{-8}. That would mean that SNPs, if selected at random, would have a one in hundred million probability of showing the effect found under the rejected hypothesis. If you have 10^{20} SNPs, 10^{12} SNPs should show such an effect. Instead, you have, per utu above, about 10^{3,000,000} (2^{10,000,000}) SNPs. Even if only one in a trillion (10^{12}) SNPs vary by ethnicity, this result remains.

    If you are selecting SNPs based on their apparent effect, the p-score is no longer indicative of improbability, as you are choosing among those that show an apparent effect, i.e. the p-score applies to the total set of SNPs, but a priori not to the subset selected, irrespective of whether causality is present. If a small number (much less than 1/p) of potential explanatory variables exist, p is impressive.

    As such, p-score and the like are irrelevant for demonstrating strength of causal probability. As causality may be present, an obvious test is to check for the consequences of lack of causality, namely a growing number of explanatory variables to maintain (and increase) R^2. For a constant R^2, does the number of explanatory SNPs increase with the size of the dataset, or not?
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  123. @res
    Simple question. Do you understand what "multiple hypothesis testing" means?

    If so, do you see how it applies to choosing subsets of SNPs for a PGS?

    If not, this might help: https://en.wikipedia.org/wiki/Multiple_comparisons_problem

    I read the article. You choose a number of hypotheses (SNPs being significant for IQ), and find an estimate of the number of erroneous “discoveries” (SNPs that appear to contribute but do not, based on randomness), and erroneous rejections (SNPs that appear to be irrelevant to IQ but in fact contribute, again due to randomness), both under the hypothesis of a per-SNP test error rate (alpha).

    That begs the question of the basis for estimation of such an error rate. A further hypothesis (perhaps constrained by real data) is needed. Do you wish to posit an explicit hypothesis, perhaps with falsification criteria?

    Read More
    • Replies: @res
    Do you understand the issue with combinatorial explosion of the possible subsets? If you choose a subset of k SNPs from n possible IQ SNPs you need to correct for n choose k possible hypotheses: https://en.wikipedia.org/wiki/Combination

    The point of my comment 117 was that it is better to avoid the need to correct for multiple hypotheses by either using all of the SNPs (rather than a subset) or defining a consistent subset criteria to use (e.g. 10 SNPs with smallest p-values) before evaluating the result.

    I get the sense that we are not understanding each other so I probably won't continue this part of the conversation.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  124. utu says:
    @Johan Meyer
    A simple hypothesis may explain the discrepancy between group (aggregate) and individual effects, to wit, genetically variable sensitivity to environment, with shared aggregate environment being negatively correlated with the genes varying sensitivity to environment. Thus more sensitive individuals may be more concentrated in places where the pathological environment is less of an issue.

    Other hypotheses could also explain such scenarios, e.g. additive hypotheses with missing variables, or failure to include individual-level second moments (covariance matrices).

    The problem is precisely the lack of adequate specification of hypotheses.

    A simple hypothesis may explain

    You seem to overuse the word “simple”. Nothing is really simple in this problem. The mathematical problem is insanely undetermined. There are 2^10,000,000 of potential solutions. Steven Hsu reduced the size of the problem by looking for solutions within a 50,000 only subset of SNPs where each SNP in the subset had a detectable correlation with the trait. Then he applied L1-fit (Lasso method) to find the smallest subset among them that maximized the explained variance. He ended up with 10,000 SNPs in his linear polygenic score. But he did not prove this was the only solution. It is very likely there are solutions among 10,000,000-50k subset that he did not use not mentioning the nonlinear polygenic score solutions that perhaps may require less than 10,000 SNPs to achieve a fit of the same quality.

    Returning to the issue of averages of subsets. If you have two variables Y an X (long lists of numbers) they may not correlate with each other but you may subdivide them into subsets X_i, Y_i, where i=0,1,2,… and the averages and of the subsets may correlate with each other. This fact tells us nothing about the correlation or lack of it between X and Y.

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  125. res says:
    @Johan Meyer
    I read the article. You choose a number of hypotheses (SNPs being significant for IQ), and find an estimate of the number of erroneous "discoveries" (SNPs that appear to contribute but do not, based on randomness), and erroneous rejections (SNPs that appear to be irrelevant to IQ but in fact contribute, again due to randomness), both under the hypothesis of a per-SNP test error rate (alpha).

    That begs the question of the basis for estimation of such an error rate. A further hypothesis (perhaps constrained by real data) is needed. Do you wish to posit an explicit hypothesis, perhaps with falsification criteria?

    Do you understand the issue with combinatorial explosion of the possible subsets? If you choose a subset of k SNPs from n possible IQ SNPs you need to correct for n choose k possible hypotheses: https://en.wikipedia.org/wiki/Combination

    The point of my comment 117 was that it is better to avoid the need to correct for multiple hypotheses by either using all of the SNPs (rather than a subset) or defining a consistent subset criteria to use (e.g. 10 SNPs with smallest p-values) before evaluating the result.

    I get the sense that we are not understanding each other so I probably won’t continue this part of the conversation.

    Read More
    • Replies: @Johan Meyer
    Rather than "testing" the hypotheses (laughable, considering the expected number of false positives that may exist but were not necessarily included), pick a subset for test. If you prefer, you could choose among those that maintain a small apparent p-value.

    Within that subset of possibilities, it should be possible to find, within 48 hours of runtime, a smallest subset that meets the needed R^2s. If the p-scores are indicative of causality, a small set should exist, with low p-values, that can meet the R^2 requirements, that should not grow much as the data size grows.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  126. @res
    Simple question. Do you understand what "multiple hypothesis testing" means?

    If so, do you see how it applies to choosing subsets of SNPs for a PGS?

    If not, this might help: https://en.wikipedia.org/wiki/Multiple_comparisons_problem

    I should also add the following. Assume a p score of 10^{-8}. That would mean that SNPs, if selected at random, would have a one in hundred million probability of showing the effect found under the rejected hypothesis. If you have 10^{20} SNPs, 10^{12} SNPs should show such an effect. Instead, you have, per utu above, about 10^{3,000,000} (2^{10,000,000}) SNPs. Even if only one in a trillion (10^{12}) SNPs vary by ethnicity, this result remains.

    If you are selecting SNPs based on their apparent effect, the p-score is no longer indicative of improbability, as you are choosing among those that show an apparent effect, i.e. the p-score applies to the total set of SNPs, but a priori not to the subset selected, irrespective of whether causality is present. If a small number (much less than 1/p) of potential explanatory variables exist, p is impressive.

    As such, p-score and the like are irrelevant for demonstrating strength of causal probability. As causality may be present, an obvious test is to check for the consequences of lack of causality, namely a growing number of explanatory variables to maintain (and increase) R^2. For a constant R^2, does the number of explanatory SNPs increase with the size of the dataset, or not?

    Read More
    • Replies: @utu

    per utu above, about 10^{3,000,000} (2^{10,000,000}) SNPs.
     
    There are 10,000,000 SNPs. There are 2^10,000,000 subsets of SNPs.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  127. @res
    Do you understand the issue with combinatorial explosion of the possible subsets? If you choose a subset of k SNPs from n possible IQ SNPs you need to correct for n choose k possible hypotheses: https://en.wikipedia.org/wiki/Combination

    The point of my comment 117 was that it is better to avoid the need to correct for multiple hypotheses by either using all of the SNPs (rather than a subset) or defining a consistent subset criteria to use (e.g. 10 SNPs with smallest p-values) before evaluating the result.

    I get the sense that we are not understanding each other so I probably won't continue this part of the conversation.

    Rather than “testing” the hypotheses (laughable, considering the expected number of false positives that may exist but were not necessarily included), pick a subset for test. If you prefer, you could choose among those that maintain a small apparent p-value.

    Within that subset of possibilities, it should be possible to find, within 48 hours of runtime, a smallest subset that meets the needed R^2s. If the p-scores are indicative of causality, a small set should exist, with low p-values, that can meet the R^2 requirements, that should not grow much as the data size grows.

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  128. utu says:
    @Johan Meyer
    I should also add the following. Assume a p score of 10^{-8}. That would mean that SNPs, if selected at random, would have a one in hundred million probability of showing the effect found under the rejected hypothesis. If you have 10^{20} SNPs, 10^{12} SNPs should show such an effect. Instead, you have, per utu above, about 10^{3,000,000} (2^{10,000,000}) SNPs. Even if only one in a trillion (10^{12}) SNPs vary by ethnicity, this result remains.

    If you are selecting SNPs based on their apparent effect, the p-score is no longer indicative of improbability, as you are choosing among those that show an apparent effect, i.e. the p-score applies to the total set of SNPs, but a priori not to the subset selected, irrespective of whether causality is present. If a small number (much less than 1/p) of potential explanatory variables exist, p is impressive.

    As such, p-score and the like are irrelevant for demonstrating strength of causal probability. As causality may be present, an obvious test is to check for the consequences of lack of causality, namely a growing number of explanatory variables to maintain (and increase) R^2. For a constant R^2, does the number of explanatory SNPs increase with the size of the dataset, or not?

    per utu above, about 10^{3,000,000} (2^{10,000,000}) SNPs.

    There are 10,000,000 SNPs. There are 2^10,000,000 subsets of SNPs.

    Read More
    • Replies: @Johan Meyer
    Thanks for the correctional. My apologies to res.
    , @Factorize
    Thinking about this in terms of the possible combinations would make this hopelessly confusing. Another possibly more helpful way is to think of the county Fair. The coin can fall down the peg board in a large number of ways though there are only a handful of ultimate bins that it can fall into. Same with the PGS. Each person in the GWAS can either have 0, 1 or 2 of the effect alleles and would score 0 for all the non effect SNPs. After a massive number of possible combinations people are placed into a small number of PGS bins. In fact for the most part the education system places people into the high school or college bin, even though there is substantial differences in genetic potential. To more fully capitalize on existing genetic potential in the community perhaps each year of education beyond high school should be considered a graduation.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  129. @utu

    per utu above, about 10^{3,000,000} (2^{10,000,000}) SNPs.
     
    There are 10,000,000 SNPs. There are 2^10,000,000 subsets of SNPs.

    Thanks for the correctional. My apologies to res.

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  130. Factorize says:

    Beyond the technical questions, it is useful to consider the new 21st vista that this research has revealed to us. We can now see a future in which science will give us a life saturated with intelligence. Psychometric science will finally unite and not divide us according to race or class. The symbiosis that has emerged within the class structure will quickly fade and with it much of the economic transfers that helps perpetuate such social distinctions.

    It should rapidly become obvious that we can not afford not to have the maximal realization of intellectual potential. The technological progress (propelled by high cognitive ability) that has occurred over the last several decades has greatly improved our lives. We have access to very affordable and highly powerful computer, genomic and other technologies. Why would we want to hold back this engine of prosperity? As soon as national governments establish programs that will allow for genetic enhancement of their entire population, we will enter an entirely new phase in humanity’s journey! Let’s get started! A life of ubiquitous genius will be a life of overwhelming accomplishment. We are so blessed that we will have the opportunity to watch this extraordinary future unfold.

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter Display All Comments
  131. j2 says:
    @utu
    You would not ave to write things 2 or 3 times if you knew that that it was unnecessary writing even the first time.

    Comment #17
    The problem is complex. When using linear polygenic score it becomes much simpler mathematically. But there are simplifications which I am not sure that are justified. Presumably they might be justified by GWAS. I do not know. For example a linear polygenic score implies that the additive effect from 0 allele to 1 allele change is the same as from 1 allele two 2 alleles. Then it implies that different SNPs are ignorant of each other while it might be possible that the effect of having SNP1 and SNP2 is not a sum of effects of the two SNPs together. To discover it one would hav to either do non-linear polygenic score or even better to an equivalent of logistic regression where SNPs cannot be mathematized to numerical values.
     
    So obviously when I used the term polygenic score in arguing a more general point its linearity was not implied.

    This is the third time I write this, will not write a fourth time.
     
    What was it you were trying to explain?

    “What was it you were trying to explain?”

    Point taken, so I try to write something more reasonable to better address your comment. To get the polygenic score method more predictive you have to find SNP combinations that effect IQ. Separate from the samples high and low performance subsets and try to identify with 2 SNP or 3 SNP combinations appear more often. Then continue this way as far as it goes, maybe to 4 SNP combinations.

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  132. hyperbola says:
    @utu
    I am looking at the problem form the mathematical end. I am not really interested in mechanism what SNPs do or not. It appears that nobody really does know what exactly SPNs form the noncoding regions do but everybody agrees that they are important because they must be included in polygenic scores to give the scores predictive power.

    I do not know how phone numbers associated with different people made those people to move to particular geographic locations or was it the others way around. I do not care what is the direction of the causality vector. But having a sufficiently large sample of people, their phone numbers and their geographic locations I can discover the rule of this association. And if I can confirm the rule on an independent validation sample I will be able to decode (lat,long) from phone numbers.

    This is really the bottom line of IQ nature-nurture controversy. What is the maximal prediction power with any polygenic score? Is it close to heritability based on twin studies or not. How this or that SNP make somebody smarter or not is secondary to me.

    As for 10 million SNPs, this study only measured 1 million people. Can we expect 7000 times more SNPs when we expand to 7 billion humans?
     
    What are you talking about? Things do not scale like that.

    I am looking at the problem form the mathematical end. I am not really interested in mechanism what SNPs do or not. It appears that nobody really does know what exactly SPNs form the noncoding regions do but everybody agrees that they are important because they must be included in polygenic scores to give the scores predictive power.

    Not a wise attitude. There is already evidence that some regions of DNA are NOT under evolutionary selection, that is, SNPs in such regions may be completely random and insignificant for any trait. The very low levels of sampling currently available and the necessity of thousands of SNPs to get any “predictive” power (even as an average), suggest that much of this “predictive power” is simply noise.

    As for “scaling of SNPs”, since for a million people we have only measured about 1 part per 7000 of human DNA (and that part is highly weighted towards certain populations), I doubt anyone knows how the scaling goes.

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  133. Okechukwu says:
    @res

    There is no selection bias in African immigration. Immigrants represent a cross-section of African society and not some elite.
     
    LOL! Here we go with this one again. Let's take a look at the IAB Brain Drain data for immigration once more. I have shown you this enough times it is clear you are just a liar.

    https://www.iab.de/en/daten/iab-brain-drain-data.aspx

    The 2010 emigration rates (men and women combined) from Nigeria by educational level were:
    Total   | Low    | Medium | High
    0.61% | 0.11% | 0.56%     | 12.04%

    Nope, not selective at all. Just a 100:1 ratio of emigration between the high and low groups.

    It is fun to note that the credentials brought by African immigrants to the US which you trumpet so loudly as proof of excellence is exactly what disproves your statement about immigration selectivity.

    Besides, basic common sense should tell you that you cannot get an elite capable of grossly outperforming whites from a source population of 55 IQ retards.
     
    So it's 55 IQ now? You should not speak so ill of the country of your ancestral origin. Even your hated Lynn did not give an estimate that low. Although his 69 for Nigeria does seem rather pessimistic: https://pdfs.semanticscholar.org/48d7/843f6ce714a684a93530a0c8b7da65d185db.pdf

    LOL! Here we go with this one again. Let’s take a look at the IAB Brain Drain data for immigration once more. I have shown you this enough times it is clear you are just a liar.

    Brain drain pertains to educated professionals, not necessarily to the most intelligent. But even as regards educated professionals, brain drain is largely a myth. The supply of educated Nigerians is not finite. They have a robust post-secondary system out of which most graduates stay in Nigeria. Also, many return to Nigeria after their academic work is completed in the west. Particularly with Igbos, the idea of being buried in a foreign land is anathema. If they don’t make it while alive, their bodies are certainly sent back if they die here.

    The smartest Nigerians are actually in Nigeria. Immigrating to the United States is extremely challenging and most don’t have the means, the wherewithal or the family connections to do so. It only seems like the Nigerians in America are elite because America compares very favorably to Nigeria in terms of the ability to achieve by dint of hard work and perseverance. Having been exposed to a hardscrabble, difficult experience in Nigeria, America becomes a relatively easy downhill ride.

    It is fun to note that the credentials brought by African immigrants to the US which you trumpet so loudly as proof of excellence is exactly what disproves your statement about immigration selectivity.

    Those credentials are acquired in the United States, otherwise they would be worthless. All references to the educational attainment of Africans in America refer exclusively to degrees acquired at American institutions. This is the point I’m trying to get across to you. Most Nigerian immigrants don’t arrive as elites. They become elites in the United States.

    So it’s 55 IQ now? You should not speak so ill of the country of your ancestral origin. Even your hated Lynn did not give an estimate that low. Although his 69 for Nigeria does seem rather pessimistic:

    But on the map whose veracity you seem to be vouching for, some African countries are assigned IQ’s that low. And even 69 would meet the standard for severe mental disability. That would mean that all Nigerian Olympic athletes would be shifted to the Special Olympics. It would also mean that Nigerians probably couldn’t operate an automobile or brush their own teeth. So racist pseudoscientists like Lynn were hoisted by their own petard. Believing that their audience wouldn’t extend beyond white nationalist simpletons, they went much too far in their mission to establish the intellectual inferiority of Africans. In fact due to negative peer reviews, Lynn has yet again revised his IQ estimate for Nigeria, this time to 84. He basically just pulls numbers out of his ass.

    https://iq-research.info/en/average-iq-by-country

    Furthermore, according to Lynn, Sierra Leone with an IQ of 91 scores higher than a host of European, Latin American, Middle Eastern and Asian countries. Kind of runs counter to the HBD narrative, doesn’t it?

    Here’s the thing, anyone that tells you they have credible IQ scores for any African country is lying. No valid, properly controlled, large sample size survey has ever been done. In most cases the scores were simply extrapolated against a backdrop of the researcher’s own deeply ingrained biases. At other times they would walk into a village and test a few illiterate, malnourished kids and on that basis arrive at an average for the country as a whole, making sure of course to weed out all high scores as anomalous. With this kind of methodology I could get an average IQ of 50 for white Americans. I could simply test attendees at a Trump rally or I could test the inhabitants of backwoods Appalachia, making sure to throw out any scores above 50.

    Read More
    • LOL: res
    • Replies: @James Thompson
    Here is a recent, large scale study of intelligence in Nigeria.
    http://www.unz.com/jthompson/sex-differences-in-intelligence-in-nigeria
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  134. @Okechukwu

    LOL! Here we go with this one again. Let’s take a look at the IAB Brain Drain data for immigration once more. I have shown you this enough times it is clear you are just a liar.
     
    Brain drain pertains to educated professionals, not necessarily to the most intelligent. But even as regards educated professionals, brain drain is largely a myth. The supply of educated Nigerians is not finite. They have a robust post-secondary system out of which most graduates stay in Nigeria. Also, many return to Nigeria after their academic work is completed in the west. Particularly with Igbos, the idea of being buried in a foreign land is anathema. If they don't make it while alive, their bodies are certainly sent back if they die here.

    The smartest Nigerians are actually in Nigeria. Immigrating to the United States is extremely challenging and most don't have the means, the wherewithal or the family connections to do so. It only seems like the Nigerians in America are elite because America compares very favorably to Nigeria in terms of the ability to achieve by dint of hard work and perseverance. Having been exposed to a hardscrabble, difficult experience in Nigeria, America becomes a relatively easy downhill ride.


    It is fun to note that the credentials brought by African immigrants to the US which you trumpet so loudly as proof of excellence is exactly what disproves your statement about immigration selectivity.
     
    Those credentials are acquired in the United States, otherwise they would be worthless. All references to the educational attainment of Africans in America refer exclusively to degrees acquired at American institutions. This is the point I'm trying to get across to you. Most Nigerian immigrants don't arrive as elites. They become elites in the United States.

    So it’s 55 IQ now? You should not speak so ill of the country of your ancestral origin. Even your hated Lynn did not give an estimate that low. Although his 69 for Nigeria does seem rather pessimistic:
     
    But on the map whose veracity you seem to be vouching for, some African countries are assigned IQ's that low. And even 69 would meet the standard for severe mental disability. That would mean that all Nigerian Olympic athletes would be shifted to the Special Olympics. It would also mean that Nigerians probably couldn't operate an automobile or brush their own teeth. So racist pseudoscientists like Lynn were hoisted by their own petard. Believing that their audience wouldn't extend beyond white nationalist simpletons, they went much too far in their mission to establish the intellectual inferiority of Africans. In fact due to negative peer reviews, Lynn has yet again revised his IQ estimate for Nigeria, this time to 84. He basically just pulls numbers out of his ass.

    https://iq-research.info/en/average-iq-by-country

    Furthermore, according to Lynn, Sierra Leone with an IQ of 91 scores higher than a host of European, Latin American, Middle Eastern and Asian countries. Kind of runs counter to the HBD narrative, doesn't it?

    Here's the thing, anyone that tells you they have credible IQ scores for any African country is lying. No valid, properly controlled, large sample size survey has ever been done. In most cases the scores were simply extrapolated against a backdrop of the researcher's own deeply ingrained biases. At other times they would walk into a village and test a few illiterate, malnourished kids and on that basis arrive at an average for the country as a whole, making sure of course to weed out all high scores as anomalous. With this kind of methodology I could get an average IQ of 50 for white Americans. I could simply test attendees at a Trump rally or I could test the inhabitants of backwoods Appalachia, making sure to throw out any scores above 50.

    Here is a recent, large scale study of intelligence in Nigeria.

    http://www.unz.com/jthompson/sex-differences-in-intelligence-in-nigeria

    Read More
    • Replies: @Okechukwu

    Here is a recent, large scale study of intelligence in Nigeria.
     

    Here is a recent, large scale study of intelligence in Nigeria.
     
    I am very skeptical of this report in light of the dubious persons and organizations involved: Richard Lynn, Mankind Quarterly and the Pioneer Fund vehicle Ulster Institute. I would hope that serious, non-ideological researchers would audit it for authenticity and genuity.

    This definite sample comes up with a result of IQ 70 and I think should be considered the best estimate of Nigerian IQ.
     
    This is deeply problematic. To suggest that perfectly functional people are, in effect, functionally retarded is to give ammunition to those who assert, with abundant justification, that IQ tests don’t actually test for intelligence. You’re never in this universe going to convince anyone that Nigerian kids are 30 points less intelligent that kids anywhere in the world. You’ll convince a small fringe of true believer trolls like res. But 99.999999% of the world will laugh at the absurdity.

    This isn’t the 19th century. We have data gleaned from the real world under real conditions that make a mockery of your assertions. Elementary, middle and high school-aged Nigerian kids do immigrate to the United States. Far from demonstrating a reduced capacity, the Nigerians often will outclass their white American counterparts. These are the same “retarded” kids from the same schools this alleged study was conducted in. When they arrive in America they’re often a few grades ahead. They marvel at how easy it is at American schools, having been exposed to a more rigorous curricula in Nigeria. Unlike Nigeria, in America the teachers are friendly, the parents are indulgent, the kids are out of control and the tests are open book.

    Don’t take my word for it. Here are some Nigerian kids talking about their experiences at American schools.

    https://www.naijarules.com/index.php?threads/american-high-school-vs-nigerian-high-school.6393/
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  135. Chainsaw1 says:
    @Johan Meyer
    Your bad language fails to hide your bad argument. I shall spell out your logical failure, to wit, why the issue remains when considering a population. Note that this is at most grade 12 probability.

    Let p_\ell be the frequency of allele \ell. Then q_\ell is the frequency of its absence. These probabilities sum to unity. There are four possibilities:

    Allele is doubly absent (probability is q_\ell^2).

    Allele is present in first carrying chromosomal homologue, but absent in second. Or allele is present in second, but missing in first. The probability of each is p_\ell q_\ell, for a total of 2p_\ell q_\ell. For low frequencies of the allele, this may be approximated as 2p_\ell, and for high frequencies, as 2q_\ell=2(1-p_\ell). Thus one may expect to find reasonable correlations away from middle grounds.

    Allele is present in both homologues (probability is p_\ell^2).

    It should not be too hard to redo the math replacing (p)robability with pq, i.e. p(1-p), and including a p^2 term with its own coefficient to be found.

    All of which is aside to whether the genes add directly to intelligence, or merely modulate the uptake of neurotoxins, which would produce the same twin correlation structure as would genes adding directly to IQ.

    You are unnecessary complicate thing by putting in all that extra variable and subscript, using procedures that do not conform to the format of the data available. In PGS effective allele frequncy (EAF) is the frequency of the effective allele in the dataset and it is already given in decimal fraction frequency, e.g.

    SNP,A1,A2,EAF,Beta,SE,Pval
    rs9859556,T,G,0.6905,0.029,0.001,3.98E-91

    where A1 is the effective allele and Beta is the effect size for the sample. They do not care about q which has zero, zilch, nada direct effect on the PGS score. And they are not interested in doing algebraic manipulation, what they want is to have minimum prediction error over the whole sample. So all your algebraic manipulations are for freaking zilch, nothing.

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  136. Chainsaw1 says:
    @Johan Meyer
    The criteria are achieving the correlations (R or R^2, as previously achieved) of 0.7, 0.8 and 0.9.

    One cannot decrease R^2 by adding explanatory variables. Thus, the question is whether the added explanatory variables used to fit a bigger data set are because of greater insight offered by the additional data, or whether it is simply finding a random pattern that happens to match tolerably well.

    If understanding is truely advancing, as suggested by increasing R^2 over time, then one should expect that lesser R^2s can be achieved by a small set, even as the data grows with time.

    Thus the challenge is to find the smallest set (preferably with the same coefficients/betas as the study being published) that will yield R^2s of at least 0.7, 0.8, and 0.9.

    If the sizes of these minimal R^2 achieving sets grow as data is added, it raises the concern that one is achieving better R^2s by inflating the explanatory variables from a random set, rather than achieving understanding.

    My suspicion is that polygenic scores are real, and that the set sizes will not grow for sets taken prior to lead clean up, as I suspect that it is a polygenic score for uptake of lead. If the sizes fail to grow even after lead clean up, it would tend to point to direct genetic contribution to IQ.

    Now learn your stats 102.

    https://en.wikipedia.org/wiki/Coefficient_of_determination#Adjusted_R2

    “If the sizes of these minimal R^2 achieving sets grow as data is added, it raises the concern that one is achieving better R^2s by inflating the explanatory variables from a random set, rather than achieving understanding. Unlike R2, the adjusted R2 increases only when the increase in R2 (due to the inclusion of a new explanatory variable) is more than one would expect to see by chance.

    Now for the stats 103.

    https://en.wikipedia.org/wiki/Akaike_information_criterion

    “AIC rewards goodness of fit (as assessed by the likelihood function), but it also includes a penalty that is an increasing function of the number of estimated parameters. The penalty discourages overfitting, because increasing the number of parameters in the model almost always improves the goodness of the fit.”

    Read More
    • Replies: @Johan Meyer
    With a large enough collection of random variables, for a given data size (n), one may find a subset of the random variables that achieve a given goodness of fit. Note the presence of n in the adjusted R^2. Note also that p is the number of variables used in a given model, not the number from which one gets to choose in making a model. If n is much larger than p, the penalty for adding new explanatory variables is small. Thus adjusted R^2 does not address the issue of selecting a subset of explanatory variables from a very large set of non-causal random variables.

    AIC uses log likelihood to discount the penalty of adding explanatory variables. The likelihood in question is the likelihood that the data would occur given the hypotheses used. Again, with a sufficient number of random variables, one may find a subset that would have high likelihood of producing the found IQ data.

    Thus both your links are irrelevant.

    Rather, the problem was with taking 2^10,000,000 instead of 10,000,000, which renders my objection to res irrelevant.

    An estimate is needed of the number of SNPs that vary by ethnicity. P-scores should by multiplied by that number, IMAO, to account for the likelihood of selecting at random SNPs that happens to fit the IQ data due to random chance, although this applies as well to other fields that have a large pool of potential explanatory variables, akin to the SNP problem.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  137. Sparkon says:

    Sorry to arrive at this late date to point out the turd in the gene pool, but evidence is accumulating that we are shaped and controlled to some degree by the microbes in our intestines.

    But these microbes’ reach may extend much further, into the human brains. A growing group of researchers around the world are [sic] investigating how the microbiome, as this bacterial ecosystem is known, regulates how people think and feel. Scientists have found evidence that this assemblage—about a thousand different species of bacteria, trillions of cells that together weigh between one and three pounds—could play a crucial role in autism, anxiety, depression, and other disorders.

    When Gut Bacteria Change Brain Function

    ["growing group... is investigating"]

    ‘Even more direct evidence here that the microbiome plays a significant role in intelligence.

    Scientists may be able to use the bacteria living in your baby’s poop to predict how he or she will perform on cognitive tests by age two, a new study suggests. Researchers took fecal samples from nearly 100 one-year-olds and found that those with less diverse microbiomes and higher levels of the bacterial genus Bacteroides had higher cognitive scores one year later.

    Rebecca Knickmeyer: “This is the first time an association between microbial communities and cognitive development has been demonstrated in humans.”

    Scientists Analyzed Babies’ Poop To Predict How Smart They’d Be

    Therefore, it is conceivable that an individual’s IQ may be determined largely by the bacteria living in his gut — the microbiome. It is possible too that geographical IQ differences may be explained by the various populations of intestinal bacteria to which inhabitants of the various geographical regions are exposed. Note that people who live with dogs share gut bacteria with that animal, so it is established that populations of gut bacteria can be affected by invaders from the animal kingdom.

    I suggest that a person’s intelligence and behavior will be affected — if not entirely controlled — by his microbiome.

    Perhaps Dr. Thompson will explore this fascinating topic in a future article.

    Read More
    • Replies: @James Thompson
    As the field develops it will be worth looking at it. No reason a priori why it should not have an effect, but I currently doubt it will be a major factor. Certainly worth checking.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  138. Okechukwu says:
    @James Thompson
    Here is a recent, large scale study of intelligence in Nigeria.
    http://www.unz.com/jthompson/sex-differences-in-intelligence-in-nigeria

    Here is a recent, large scale study of intelligence in Nigeria.

    Here is a recent, large scale study of intelligence in Nigeria.

    I am very skeptical of this report in light of the dubious persons and organizations involved: Richard Lynn, Mankind Quarterly and the Pioneer Fund vehicle Ulster Institute. I would hope that serious, non-ideological researchers would audit it for authenticity and genuity.

    This definite sample comes up with a result of IQ 70 and I think should be considered the best estimate of Nigerian IQ.

    This is deeply problematic. To suggest that perfectly functional people are, in effect, functionally retarded is to give ammunition to those who assert, with abundant justification, that IQ tests don’t actually test for intelligence. You’re never in this universe going to convince anyone that Nigerian kids are 30 points less intelligent that kids anywhere in the world. You’ll convince a small fringe of true believer trolls like res. But 99.999999% of the world will laugh at the absurdity.

    This isn’t the 19th century. We have data gleaned from the real world under real conditions that make a mockery of your assertions. Elementary, middle and high school-aged Nigerian kids do immigrate to the United States. Far from demonstrating a reduced capacity, the Nigerians often will outclass their white American counterparts. These are the same “retarded” kids from the same schools this alleged study was conducted in. When they arrive in America they’re often a few grades ahead. They marvel at how easy it is at American schools, having been exposed to a more rigorous curricula in Nigeria. Unlike Nigeria, in America the teachers are friendly, the parents are indulgent, the kids are out of control and the tests are open book.

    Don’t take my word for it. Here are some Nigerian kids talking about their experiences at American schools.

    https://www.naijarules.com/index.php?threads/american-high-school-vs-nigerian-high-school.6393/

    Read More
    • Replies: @James Thompson
    Agree we should always look for anomalies between tests and achievements, and that real life achievements are the real arbiter. However, there is a real puzzle here. The intelligence test results in Africa are very low. The academic achievement scores in Africa are very low. Most African economic indicators are low, bar extraction of raw materials from mines.
    Two arguments you will find in my debates with Chandra Chisala: 1) populations are very large in Africa, so even if intelligence results are correct there are some outliers 2) African populations vary, and some may be cognitive elites.
    Kierkegaard has looked at whether African are better at "mental sports" competitions, and finds they are not. Could be lack of internet access.
    At the moment we don't, in my view, have a resolution of this, other than saying that the results, however alarming, seem to fit together.
    Contrary data are welcome.
    , @APilgrim
    The unrelenting & witless 'racial-IQ-correlation' denialism of Okechukwu clearly indicates that the fellow is thick & black.

    But, I repeat myself.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  139. Edward says:

    Brilliant to see that we’re really starting to dig into the neurobiological and physiological basis of g. The next few decades of work in this area will be exhilarating.

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter Display All Comments
  140. Factorize says:
    @utu

    the normal distribution for IQ arose from empirical observations
     
    It appears to be so but do not forget that the IQ scale is kind of arbitrary. It was constructed. And I would not be surprised if obtaining Gaussian distribution was the objective of the engineers.

    If your n-th question in your test is answered by only 1.5% of population you will know what weight to assign to that question in the final score to be on the track to Gaussian distribution construction.

    And in the end you can always perform a nonlinear scale adjustment. Any concave pfd(IQ) distribution can be transformed into a Gaussian one by some monotonic f(IQ)-->IQ transformation.

    The bottom line is that whether IQ score is Gaussian or not is unimportant. But on the other hand what kind of distributions can be constructed using a polygenic score might be more interesting.

    utu, I have been considering your comments about the distribution of IQ. It occurred to me that you could force the distribution to nearly whatever you liked simply by selecting certain questions. For example the distribution of cognitive ability would seem to be very different if the one and only question were to derive E= mc2 from first principles instead of 1 + 1 =? Yet, what you are really trying to do with an IQ test is to provide questions of increasing complexity that will pick up differences in ability levels. The two questions above really do not achieve this objective. With 1 + 1, presumably most would not need a calculator. One could then gradually escalate the complexity involved. Perhaps at the next graduation, one might ask what 1 +
    4 = and so on. The complexity level could then continue to escalate. At the top level, perhaps you could provide a 15 digit number and ask for the 8th root.
    When this construction of IQ test is used a normal distribution results.
    Starting at the low end, more and more people would be able to answer easy questions such as 7+8, until you reached a point where average people might be challenged for example 17*31. From this level forward, fewer and fewer would be able to provide the correct answer.

    One counterpoint that I do find troubling is that one of our teachers became quite distressed that the normal curve did not seem to apply in her class. She was called into the principal’s office to explain this discrepancy. In her defence, she noted that the Asian students had scored so high that the entire distribution was non-normal. I do not feel comfortable with such a dogmatic stance: you really shouldn’t be expected to provide the expected empirical answer when your observed answer is different. In a science course that would be simply fixing your labs. If enough people buy into the group think then the answer that you wind up with is simply a self-fulfilling prophecy.

    One aspect of the EA3 paper that has strangely wentunmentioned is that this GWAS has now brought this new EA research era to main Street. Let’s face it, there are likely only a few global organizations that could meaningfully contribute to a 1.1 million person GWAS. It is now a quite small and intimate closet full of people doing this science.
    Now think of who might have the resources to do a 75 person study. This size was noted in the paper as being able to confirm the EA3 results. Basically, almost any organization could scrounge up the money necessary to do such a study. Probably even high schools would have the funds needed for their students to do this research.

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  141. Factorize says:

    utu, thank you for responding!

    I thought about your comments about IQ distribution. I realize that the test can largely shape the distribution. Consider IQ tests based on either the question 1 + 1 = ? or Derive from first principles E= mc2.
    The psychometric design problem with these tests is that they are totally obscuring individual differences by making the test either too easy or too hard. A well constructed test would have questions or various levels of complexity so that individual differences could become obvious. So instead of 1 + 1 as the test, there could be a number of intermediate level questions with perhaps a final question such as find the eighth root of a 15 digit number. Constructing the test in this way would produce a normal distribution for cognitive ability

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter Display All Comments
  142. Factorize says:
    @utu

    per utu above, about 10^{3,000,000} (2^{10,000,000}) SNPs.
     
    There are 10,000,000 SNPs. There are 2^10,000,000 subsets of SNPs.

    Thinking about this in terms of the possible combinations would make this hopelessly confusing. Another possibly more helpful way is to think of the county Fair. The coin can fall down the peg board in a large number of ways though there are only a handful of ultimate bins that it can fall into. Same with the PGS. Each person in the GWAS can either have 0, 1 or 2 of the effect alleles and would score 0 for all the non effect SNPs. After a massive number of possible combinations people are placed into a small number of PGS bins. In fact for the most part the education system places people into the high school or college bin, even though there is substantial differences in genetic potential. To more fully capitalize on existing genetic potential in the community perhaps each year of education beyond high school should be considered a graduation.

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  143. @Sparkon
    Sorry to arrive at this late date to point out the turd in the gene pool, but evidence is accumulating that we are shaped and controlled to some degree by the microbes in our intestines.

    But these microbes’ reach may extend much further, into the human brains. A growing group of researchers around the world are [sic] investigating how the microbiome, as this bacterial ecosystem is known, regulates how people think and feel. Scientists have found evidence that this assemblage—about a thousand different species of bacteria, trillions of cells that together weigh between one and three pounds—could play a crucial role in autism, anxiety, depression, and other disorders.
     
    When Gut Bacteria Change Brain Function

    ["growing group... is investigating"]

    'Even more direct evidence here that the microbiome plays a significant role in intelligence.

    Scientists may be able to use the bacteria living in your baby’s poop to predict how he or she will perform on cognitive tests by age two, a new study suggests. Researchers took fecal samples from nearly 100 one-year-olds and found that those with less diverse microbiomes and higher levels of the bacterial genus Bacteroides had higher cognitive scores one year later.

    Rebecca Knickmeyer: “This is the first time an association between microbial communities and cognitive development has been demonstrated in humans.”
     

     
    Scientists Analyzed Babies' Poop To Predict How Smart They'd Be

    Therefore, it is conceivable that an individual's IQ may be determined largely by the bacteria living in his gut -- the microbiome. It is possible too that geographical IQ differences may be explained by the various populations of intestinal bacteria to which inhabitants of the various geographical regions are exposed. Note that people who live with dogs share gut bacteria with that animal, so it is established that populations of gut bacteria can be affected by invaders from the animal kingdom.

    I suggest that a person's intelligence and behavior will be affected -- if not entirely controlled -- by his microbiome.

    Perhaps Dr. Thompson will explore this fascinating topic in a future article.

    As the field develops it will be worth looking at it. No reason a priori why it should not have an effect, but I currently doubt it will be a major factor. Certainly worth checking.

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  144. @Okechukwu

    Here is a recent, large scale study of intelligence in Nigeria.
     

    Here is a recent, large scale study of intelligence in Nigeria.
     
    I am very skeptical of this report in light of the dubious persons and organizations involved: Richard Lynn, Mankind Quarterly and the Pioneer Fund vehicle Ulster Institute. I would hope that serious, non-ideological researchers would audit it for authenticity and genuity.

    This definite sample comes up with a result of IQ 70 and I think should be considered the best estimate of Nigerian IQ.
     
    This is deeply problematic. To suggest that perfectly functional people are, in effect, functionally retarded is to give ammunition to those who assert, with abundant justification, that IQ tests don’t actually test for intelligence. You’re never in this universe going to convince anyone that Nigerian kids are 30 points less intelligent that kids anywhere in the world. You’ll convince a small fringe of true believer trolls like res. But 99.999999% of the world will laugh at the absurdity.

    This isn’t the 19th century. We have data gleaned from the real world under real conditions that make a mockery of your assertions. Elementary, middle and high school-aged Nigerian kids do immigrate to the United States. Far from demonstrating a reduced capacity, the Nigerians often will outclass their white American counterparts. These are the same “retarded” kids from the same schools this alleged study was conducted in. When they arrive in America they’re often a few grades ahead. They marvel at how easy it is at American schools, having been exposed to a more rigorous curricula in Nigeria. Unlike Nigeria, in America the teachers are friendly, the parents are indulgent, the kids are out of control and the tests are open book.

    Don’t take my word for it. Here are some Nigerian kids talking about their experiences at American schools.

    https://www.naijarules.com/index.php?threads/american-high-school-vs-nigerian-high-school.6393/

    Agree we should always look for anomalies between tests and achievements, and that real life achievements are the real arbiter. However, there is a real puzzle here. The intelligence test results in Africa are very low. The academic achievement scores in Africa are very low. Most African economic indicators are low, bar extraction of raw materials from mines.
    Two arguments you will find in my debates with Chandra Chisala: 1) populations are very large in Africa, so even if intelligence results are correct there are some outliers 2) African populations vary, and some may be cognitive elites.
    Kierkegaard has looked at whether African are better at “mental sports” competitions, and finds they are not. Could be lack of internet access.
    At the moment we don’t, in my view, have a resolution of this, other than saying that the results, however alarming, seem to fit together.
    Contrary data are welcome.

    Read More
    • Replies: @Tim too
    When I first saw the IQ distribution map, I was reminded of the global distribution map of the sickle cell anemia gene and malaria distribution. I'm not suggesting any relationship, it was just a reminder.

    See the maps: https://www.nature.com/articles/ncomms1104
    , @Okechukwu

    The intelligence test results in Africa are very low.
     
    I don't believe we have sufficient credible data to make this assessment. There's a lot of fraudulent and unrepresentative data, particularly from Richard Lynn.

    Study of African IQ levels proven to be substandard

    http://www.uva.nl/en/content/news/news/2010/01/study-of-african-iq-levels-proven-to-be-substandard.html

    Controversial study of African IQ levels is 'deeply flawed'

    https://www.sciencedaily.com/releases/2010/01/100121155220.htm

    Having said that, to the extent that real IQ gaps exist between Africans and say, Americans, the obvious casual factor is cultural dissimilarity. As someone who is bicultural, I know that Africans tend to think in certain ways that may militate against the values westernized IQ tests are designed to reward. Or as James Flynn says, IQ doesn't test for intelligence as such, it tests for adaptation to western modernity.

    To get a fuller appreciation of the IQ's of people in the west, it may be a good idea to give them tests steeped in the norms and cultural references of non-western people. On such a test you would probably be the person with the low IQ.


    The academic achievement scores in Africa are very low.
     
    Not true. Scholars from Africa are among the highest achieving people in the world. That is to say, they receive their initial training at African institutions.

    Most African economic indicators are low
     
    If economic indicators indicated superior intellectual capacity all of the great economies that have risen and collapsed over the eons would still be with us. These things are fleeting and are driven by factors that have little to do with intelligence. Russia is a vast country of 150 million whites that has only been able to muster a GPD the size of New York City's. So we can eliminate being white as the key variable for economic success. In fact at various times in history, Europe was a barbaric backwater and economic basketcase.

    Most African economic indicators are low, bar extraction of raw materials from mines.
     
    The African Lion economies are among the fastest growing in the world. They have little in the way of raw materials.

    1) populations are very large in Africa, so even if intelligence results are correct there are some outliers 2) African populations vary, and some may be cognitive elites.
     
    There is no evidence that intelligent Africans are outliers or that intelligence emerges at a reduced frequency in Africa relative to anywhere else. It's an absurd suggestion if you know anything about how nature works. Nature would not arbitrarily decree that while Africans can be intelligent, fewer Africans of intelligence will exist. Either an organism has something or it doesn't. That's why there isn't a single chimpanzee that can operate a car or do algebraic equations. What you describe as outliers are simply people who've had the opportunity to actualize their potential.

    Let's examine some examples to give you a better idea as to why only a small percentage of potential African cognitive elites are able to realize their potential.

    From African Refugee Camp To The Ivy League

    http://flsentinel.com/from-african-refugee-camp-to-the-ivy-league/

    Back to school: From rural Africa to the Ivy League

    https://www.csmonitor.com/World/Africa/2012/0902/Back-to-school-From-rural-Africa-to-the-Ivy-League

    Minnesota Teen Accepted to All Eight Ivy League Schools

    https://www.nbcnews.com/news/us-news/minnesota-teen-munira-khalif-accepted-all-eight-ivy-league-schools-n338661

    From Poverty To The Ivy League: A Refugee's Story

    http://www.wbur.org/npr/112334064/story.php

    There are many such stories. The reason I focused on refugees is to give you an idea of how all of these people would've gone unrecognized as cognitive elites if just one thing had gone wrong for them. If they didn't meet the one critical person. Or if they didn't show up at some workshop. Or if one day they turned left instead of right. Each one of them would've gone into that massive pool of "intellectually deficient Africans." In reality, I think it's safe to say that such persons are the very tip of the iceberg who have managed through serendipity or just luck, to make it. So your assertion that African cognitive elites are outliers is false.


    Contrary data are welcome.
     
    I don't think so. You'll just say the data represents outliers. You will pretty much dismiss in this manner any piece of information that contradicts your worldview. It's a timeworn talking point that everyone's familiar with.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  145. @Chainsaw1
    Now learn your stats 102.

    https://en.wikipedia.org/wiki/Coefficient_of_determination#Adjusted_R2

    "If the sizes of these minimal R^2 achieving sets grow as data is added, it raises the concern that one is achieving better R^2s by inflating the explanatory variables from a random set, rather than achieving understanding. Unlike R2, the adjusted R2 increases only when the increase in R2 (due to the inclusion of a new explanatory variable) is more than one would expect to see by chance."

    Now for the stats 103.

    https://en.wikipedia.org/wiki/Akaike_information_criterion

    "AIC rewards goodness of fit (as assessed by the likelihood function), but it also includes a penalty that is an increasing function of the number of estimated parameters. The penalty discourages overfitting, because increasing the number of parameters in the model almost always improves the goodness of the fit."

    With a large enough collection of random variables, for a given data size (n), one may find a subset of the random variables that achieve a given goodness of fit. Note the presence of n in the adjusted R^2. Note also that p is the number of variables used in a given model, not the number from which one gets to choose in making a model. If n is much larger than p, the penalty for adding new explanatory variables is small. Thus adjusted R^2 does not address the issue of selecting a subset of explanatory variables from a very large set of non-causal random variables.

    AIC uses log likelihood to discount the penalty of adding explanatory variables. The likelihood in question is the likelihood that the data would occur given the hypotheses used. Again, with a sufficient number of random variables, one may find a subset that would have high likelihood of producing the found IQ data.

    Thus both your links are irrelevant.

    Rather, the problem was with taking 2^10,000,000 instead of 10,000,000, which renders my objection to res irrelevant.

    An estimate is needed of the number of SNPs that vary by ethnicity. P-scores should by multiplied by that number, IMAO, to account for the likelihood of selecting at random SNPs that happens to fit the IQ data due to random chance, although this applies as well to other fields that have a large pool of potential explanatory variables, akin to the SNP problem.

    Read More
    • Replies: @Johan Meyer
    In fairness to chainsaw, I did initially formulate the problem in terms of unadjusted R^2. The problem is that the expectation of the number of variables that randomly reach a given p-score is proportional to the number of potential random variables available, which is much more than those that are necessarily correlated with IQ, as the latter will have deviations from IQ.

    Thus in fact my original formulation was wrong, as chainsaw noted.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  146. APilgrim says:
    @Okechukwu

    Here is a recent, large scale study of intelligence in Nigeria.
     

    Here is a recent, large scale study of intelligence in Nigeria.
     
    I am very skeptical of this report in light of the dubious persons and organizations involved: Richard Lynn, Mankind Quarterly and the Pioneer Fund vehicle Ulster Institute. I would hope that serious, non-ideological researchers would audit it for authenticity and genuity.

    This definite sample comes up with a result of IQ 70 and I think should be considered the best estimate of Nigerian IQ.
     
    This is deeply problematic. To suggest that perfectly functional people are, in effect, functionally retarded is to give ammunition to those who assert, with abundant justification, that IQ tests don’t actually test for intelligence. You’re never in this universe going to convince anyone that Nigerian kids are 30 points less intelligent that kids anywhere in the world. You’ll convince a small fringe of true believer trolls like res. But 99.999999% of the world will laugh at the absurdity.

    This isn’t the 19th century. We have data gleaned from the real world under real conditions that make a mockery of your assertions. Elementary, middle and high school-aged Nigerian kids do immigrate to the United States. Far from demonstrating a reduced capacity, the Nigerians often will outclass their white American counterparts. These are the same “retarded” kids from the same schools this alleged study was conducted in. When they arrive in America they’re often a few grades ahead. They marvel at how easy it is at American schools, having been exposed to a more rigorous curricula in Nigeria. Unlike Nigeria, in America the teachers are friendly, the parents are indulgent, the kids are out of control and the tests are open book.

    Don’t take my word for it. Here are some Nigerian kids talking about their experiences at American schools.

    https://www.naijarules.com/index.php?threads/american-high-school-vs-nigerian-high-school.6393/

    The unrelenting & witless ‘racial-IQ-correlation’ denialism of Okechukwu clearly indicates that the fellow is thick & black.

    But, I repeat myself.

    Read More
    • Replies: @Okechukwu

    The unrelenting & witless ‘racial-IQ-correlation’ denialism of Okechukwu clearly indicates that the fellow is thick & black.
     
    That must be why Nigerian immigrants have an IQ of 110.

    Look, this dog won’t hunt. You don’t even believe it yourself otherwise you would let the facts speak for themselves rather than engaging in constant promotion and propaganda. And you’re not alone either because this idea is a fringe position that almost everyone rejects, including the Africans themselves.

    Africans certainly do not believe that anyone is superior to them intellectually. Often the converse is true. Africans think they’re smarter. It’s this conviction that whites are dumb that drives the yearly multi-billion dollar 419 industry. White marks are called mugus (look it up). I'm not talking about the poorly written letters either. I'm talking about top-level, ingenious, highly sophisticated operations that often take millions of dollars at a time. Many of the victims are the so-called high IQ types: Doctors, bank presidents, oil company executives, politicians, engineers, accountants, etc.

    So the "dumb" Africans are living in mansions and flying around in private jets by virtue of outsmarting and cleaning out the "smart" whites.

    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  147. @Johan Meyer
    With a large enough collection of random variables, for a given data size (n), one may find a subset of the random variables that achieve a given goodness of fit. Note the presence of n in the adjusted R^2. Note also that p is the number of variables used in a given model, not the number from which one gets to choose in making a model. If n is much larger than p, the penalty for adding new explanatory variables is small. Thus adjusted R^2 does not address the issue of selecting a subset of explanatory variables from a very large set of non-causal random variables.

    AIC uses log likelihood to discount the penalty of adding explanatory variables. The likelihood in question is the likelihood that the data would occur given the hypotheses used. Again, with a sufficient number of random variables, one may find a subset that would have high likelihood of producing the found IQ data.

    Thus both your links are irrelevant.

    Rather, the problem was with taking 2^10,000,000 instead of 10,000,000, which renders my objection to res irrelevant.

    An estimate is needed of the number of SNPs that vary by ethnicity. P-scores should by multiplied by that number, IMAO, to account for the likelihood of selecting at random SNPs that happens to fit the IQ data due to random chance, although this applies as well to other fields that have a large pool of potential explanatory variables, akin to the SNP problem.

    In fairness to chainsaw, I did initially formulate the problem in terms of unadjusted R^2. The problem is that the expectation of the number of variables that randomly reach a given p-score is proportional to the number of potential random variables available, which is much more than those that are necessarily correlated with IQ, as the latter will have deviations from IQ.

    Thus in fact my original formulation was wrong, as chainsaw noted.

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  148. Factorize says:

    Sometimes the math can become so complex that it seems there is no way to make even slight progress in putting the models into understandable terms. When the models seem beyond reach, people can try to discredit the entire result. Yet, it occurred to me that one simple way to attain at least a basic level of processing of the dataset would be to start with averaging. First find the sample average of educational attainment, then for each of the ten million SNPs find the average educational attainment of those with none, 1 or 2 of an arbitrarily chosen allele. With this information, you can start to built up the model. The sample average is x bar, to that you could add in your betas calculated above. If the sample average were 12 years and one SNP raised EA by 0.05 years (roughly 2 weeks), then the total sample had roughly 13.2 million school years and those with the SNP contributed an extra 55,000 years to the average. When thinking about the result in this way it becomes much easier to see that the small individual differences become large absolute differences when summed across 1.1 million people. With such a large sample, the signal becomes much easier to identify. This description should provide those with concerns about the validity of the research some comfort as it can be seen clearly above that even though the individual effect SNPs have very small effect sizes when summed across over 1 million people large aggregate EA differences emerge.

    I also started to think about how with the initial list one might then be able to create additional conditionals such as find the averages for people who had 1 of the minor allele of SNP1 and 1 of the minor allele of SNP2 etc. or building up a model by adding in one SNP at a time to the model. This became complicated so I let it go.

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter Display All Comments
  149. Tim too says:
    @James Thompson
    Agree we should always look for anomalies between tests and achievements, and that real life achievements are the real arbiter. However, there is a real puzzle here. The intelligence test results in Africa are very low. The academic achievement scores in Africa are very low. Most African economic indicators are low, bar extraction of raw materials from mines.
    Two arguments you will find in my debates with Chandra Chisala: 1) populations are very large in Africa, so even if intelligence results are correct there are some outliers 2) African populations vary, and some may be cognitive elites.
    Kierkegaard has looked at whether African are better at "mental sports" competitions, and finds they are not. Could be lack of internet access.
    At the moment we don't, in my view, have a resolution of this, other than saying that the results, however alarming, seem to fit together.
    Contrary data are welcome.

    When I first saw the IQ distribution map, I was reminded of the global distribution map of the sickle cell anemia gene and malaria distribution. I’m not suggesting any relationship, it was just a reminder.

    See the maps: https://www.nature.com/articles/ncomms1104

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  150. Okechukwu says:
    @James Thompson
    Agree we should always look for anomalies between tests and achievements, and that real life achievements are the real arbiter. However, there is a real puzzle here. The intelligence test results in Africa are very low. The academic achievement scores in Africa are very low. Most African economic indicators are low, bar extraction of raw materials from mines.
    Two arguments you will find in my debates with Chandra Chisala: 1) populations are very large in Africa, so even if intelligence results are correct there are some outliers 2) African populations vary, and some may be cognitive elites.
    Kierkegaard has looked at whether African are better at "mental sports" competitions, and finds they are not. Could be lack of internet access.
    At the moment we don't, in my view, have a resolution of this, other than saying that the results, however alarming, seem to fit together.
    Contrary data are welcome.

    The intelligence test results in Africa are very low.

    I don’t believe we have sufficient credible data to make this assessment. There’s a lot of fraudulent and unrepresentative data, particularly from Richard Lynn.

    Study of African IQ levels proven to be substandard

    http://www.uva.nl/en/content/news/news/2010/01/study-of-african-iq-levels-proven-to-be-substandard.html

    Controversial study of African IQ levels is ‘deeply flawed’

    https://www.sciencedaily.com/releases/2010/01/100121155220.htm

    Having said that, to the extent that real IQ gaps exist between Africans and say, Americans, the obvious casual factor is cultural dissimilarity. As someone who is bicultural, I know that Africans tend to think in certain ways that may militate against the values westernized IQ tests are designed to reward. Or as James Flynn says, IQ doesn’t test for intelligence as such, it tests for adaptation to western modernity.

    To get a fuller appreciation of the IQ’s of people in the west, it may be a good idea to give them tests steeped in the norms and cultural references of non-western people. On such a test you would probably be the person with the low IQ.

    The academic achievement scores in Africa are very low.

    Not true. Scholars from Africa are among the highest achieving people in the world. That is to say, they receive their initial training at African institutions.

    Most African economic indicators are low

    If economic indicators indicated superior intellectual capacity all of the great economies that have risen and collapsed over the eons would still be with us. These things are fleeting and are driven by factors that have little to do with intelligence. Russia is a vast country of 150 million whites that has only been able to muster a GPD the size of New York City’s. So we can eliminate being white as the key variable for economic success. In fact at various times in history, Europe was a barbaric backwater and economic basketcase.

    Most African economic indicators are low, bar extraction of raw materials from mines.

    The African Lion economies are among the fastest growing in the world. They have little in the way of raw materials.

    1) populations are very large in Africa, so even if intelligence results are correct there are some outliers 2) African populations vary, and some may be cognitive elites.

    There is no evidence that intelligent Africans are outliers or that intelligence emerges at a reduced frequency in Africa relative to anywhere else. It’s an absurd suggestion if you know anything about how nature works. Nature would not arbitrarily decree that while Africans can be intelligent, fewer Africans of intelligence will exist. Either an organism has something or it doesn’t. That’s why there isn’t a single chimpanzee that can operate a car or do algebraic equations. What you describe as outliers are simply people who’ve had the opportunity to actualize their potential.

    Let’s examine some examples to give you a better idea as to why only a small percentage of potential African cognitive elites are able to realize their potential.

    From African Refugee Camp To The Ivy League

    http://flsentinel.com/from-african-refugee-camp-to-the-ivy-league/

    Back to school: From rural Africa to the Ivy League

    https://www.csmonitor.com/World/Africa/2012/0902/Back-to-school-From-rural-Africa-to-the-Ivy-League

    Minnesota Teen Accepted to All Eight Ivy League Schools

    https://www.nbcnews.com/news/us-news/minnesota-teen-munira-khalif-accepted-all-eight-ivy-league-schools-n338661

    From Poverty To The Ivy League: A Refugee’s Story

    http://www.wbur.org/npr/112334064/story.php

    There are many such stories. The reason I focused on refugees is to give you an idea of how all of these people would’ve gone unrecognized as cognitive elites if just one thing had gone wrong for them. If they didn’t meet the one critical person. Or if they didn’t show up at some workshop. Or if one day they turned left instead of right. Each one of them would’ve gone into that massive pool of “intellectually deficient Africans.” In reality, I think it’s safe to say that such persons are the very tip of the iceberg who have managed through serendipity or just luck, to make it. So your assertion that African cognitive elites are outliers is false.

    Contrary data are welcome.

    I don’t think so. You’ll just say the data represents outliers. You will pretty much dismiss in this manner any piece of information that contradicts your worldview. It’s a timeworn talking point that everyone’s familiar with.

    Read More
    • Replies: @James Thompson
    Thanks for your comments and links.


    Jelte Wicherts and Richard Lynn had a series of exchanges about African intelligence. You should look at the whole exchange. The link below mentions the final conclusions they reached.

    https://www.unz.com/jthompson/james-watson-nobel-laureate-and-unperson/?highlight=Wicherts

    There is now more up to date material from Rindermann. David Becker is re-doing the Lynn database, and even after cutting out papers which have poor reporting of results, the main findings still stand.

    https://www.unz.com/jthompson/migrant-competence/?highlight=Rindermann+Africa

    If you use the search bar on my blog you will find links to many of the original papers, which is far better than newspaper reports.

    It is unusual to argue that a group of people are bright because they are dishonest, but it is certainly a hypothesis worth testing. In fact dishonesty is linked to lower ability.

    https://www.unz.com/jthompson/do-brighter-minds-incline-to-honesty/?highlight=dishonesty
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  151. Okechukwu says:
    @APilgrim
    The unrelenting & witless 'racial-IQ-correlation' denialism of Okechukwu clearly indicates that the fellow is thick & black.

    But, I repeat myself.

    The unrelenting & witless ‘racial-IQ-correlation’ denialism of Okechukwu clearly indicates that the fellow is thick & black.

    That must be why Nigerian immigrants have an IQ of 110.

    Look, this dog won’t hunt. You don’t even believe it yourself otherwise you would let the facts speak for themselves rather than engaging in constant promotion and propaganda. And you’re not alone either because this idea is a fringe position that almost everyone rejects, including the Africans themselves.

    Africans certainly do not believe that anyone is superior to them intellectually. Often the converse is true. Africans think they’re smarter. It’s this conviction that whites are dumb that drives the yearly multi-billion dollar 419 industry. White marks are called mugus (look it up). I’m not talking about the poorly written letters either. I’m talking about top-level, ingenious, highly sophisticated operations that often take millions of dollars at a time. Many of the victims are the so-called high IQ types: Doctors, bank presidents, oil company executives, politicians, engineers, accountants, etc.

    So the “dumb” Africans are living in mansions and flying around in private jets by virtue of outsmarting and cleaning out the “smart” whites.

    Read More
    • Replies: @APilgrim
    110 is an average IQ for the upper cohort (Bimodal Right Maxima), of American Caucasians.

    A cherry-picked sample of 110 IQ Nigerians represents a serious 'Brain-Drain' for Nigeria.

    We don't need Nigerian O&G. We definitely don't want any more Nigerians. British slave-traders delivered too many 4 centuries ago. They have only now been 'bred-up' to an average IQ of 85.

    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  152. @Okechukwu

    The intelligence test results in Africa are very low.
     
    I don't believe we have sufficient credible data to make this assessment. There's a lot of fraudulent and unrepresentative data, particularly from Richard Lynn.

    Study of African IQ levels proven to be substandard

    http://www.uva.nl/en/content/news/news/2010/01/study-of-african-iq-levels-proven-to-be-substandard.html

    Controversial study of African IQ levels is 'deeply flawed'

    https://www.sciencedaily.com/releases/2010/01/100121155220.htm

    Having said that, to the extent that real IQ gaps exist between Africans and say, Americans, the obvious casual factor is cultural dissimilarity. As someone who is bicultural, I know that Africans tend to think in certain ways that may militate against the values westernized IQ tests are designed to reward. Or as James Flynn says, IQ doesn't test for intelligence as such, it tests for adaptation to western modernity.

    To get a fuller appreciation of the IQ's of people in the west, it may be a good idea to give them tests steeped in the norms and cultural references of non-western people. On such a test you would probably be the person with the low IQ.


    The academic achievement scores in Africa are very low.
     
    Not true. Scholars from Africa are among the highest achieving people in the world. That is to say, they receive their initial training at African institutions.

    Most African economic indicators are low
     
    If economic indicators indicated superior intellectual capacity all of the great economies that have risen and collapsed over the eons would still be with us. These things are fleeting and are driven by factors that have little to do with intelligence. Russia is a vast country of 150 million whites that has only been able to muster a GPD the size of New York City's. So we can eliminate being white as the key variable for economic success. In fact at various times in history, Europe was a barbaric backwater and economic basketcase.

    Most African economic indicators are low, bar extraction of raw materials from mines.
     
    The African Lion economies are among the fastest growing in the world. They have little in the way of raw materials.

    1) populations are very large in Africa, so even if intelligence results are correct there are some outliers 2) African populations vary, and some may be cognitive elites.
     
    There is no evidence that intelligent Africans are outliers or that intelligence emerges at a reduced frequency in Africa relative to anywhere else. It's an absurd suggestion if you know anything about how nature works. Nature would not arbitrarily decree that while Africans can be intelligent, fewer Africans of intelligence will exist. Either an organism has something or it doesn't. That's why there isn't a single chimpanzee that can operate a car or do algebraic equations. What you describe as outliers are simply people who've had the opportunity to actualize their potential.

    Let's examine some examples to give you a better idea as to why only a small percentage of potential African cognitive elites are able to realize their potential.

    From African Refugee Camp To The Ivy League

    http://flsentinel.com/from-african-refugee-camp-to-the-ivy-league/

    Back to school: From rural Africa to the Ivy League

    https://www.csmonitor.com/World/Africa/2012/0902/Back-to-school-From-rural-Africa-to-the-Ivy-League

    Minnesota Teen Accepted to All Eight Ivy League Schools

    https://www.nbcnews.com/news/us-news/minnesota-teen-munira-khalif-accepted-all-eight-ivy-league-schools-n338661

    From Poverty To The Ivy League: A Refugee's Story

    http://www.wbur.org/npr/112334064/story.php

    There are many such stories. The reason I focused on refugees is to give you an idea of how all of these people would've gone unrecognized as cognitive elites if just one thing had gone wrong for them. If they didn't meet the one critical person. Or if they didn't show up at some workshop. Or if one day they turned left instead of right. Each one of them would've gone into that massive pool of "intellectually deficient Africans." In reality, I think it's safe to say that such persons are the very tip of the iceberg who have managed through serendipity or just luck, to make it. So your assertion that African cognitive elites are outliers is false.


    Contrary data are welcome.
     
    I don't think so. You'll just say the data represents outliers. You will pretty much dismiss in this manner any piece of information that contradicts your worldview. It's a timeworn talking point that everyone's familiar with.

    Thanks for your comments and links.

    Jelte Wicherts and Richard Lynn had a series of exchanges about African intelligence. You should look at the whole exchange. The link below mentions the final conclusions they reached.

    https://www.unz.com/jthompson/james-watson-nobel-laureate-and-unperson/?highlight=Wicherts

    There is now more up to date material from Rindermann. David Becker is re-doing the Lynn database, and even after cutting out papers which have poor reporting of results, the main findings still stand.

    https://www.unz.com/jthompson/migrant-competence/?highlight=Rindermann+Africa

    If you use the search bar on my blog you will find links to many of the original papers, which is far better than newspaper reports.

    It is unusual to argue that a group of people are bright because they are dishonest, but it is certainly a hypothesis worth testing. In fact dishonesty is linked to lower ability.

    https://www.unz.com/jthompson/do-brighter-minds-incline-to-honesty/?highlight=dishonesty

    Read More
    • Replies: @Okechukwu
    Thanks. I will take a look at the links and comment.

    It is unusual to argue that a group of people are bright because they are dishonest, but it is certainly a hypothesis worth testing. In fact dishonesty is linked to lower ability.
     
    All parties to these transactions are dishonest. The mark's dishonesty is a critical element in the successful prosecution of the scam. Indeed, their dishonesty is definitely linked to lower ability, such that they are outwitted and destroyed financially.

    Another interesting aspect of 419 is the assumption among the some of the victims that Africans are stupid and corrupt and that crimes against African countries don't carry any consequences. Imagine plotting with an American functionary to steal tens of millions of dollars from the Federal Reserve. Or plotting with Exxon-Mobil to receive oil under the table at millions of dollars below the market price. These are the kinds of illegal ventures these scam victims are engaged in with respect to Africa. Consequently, the African scam operators are able to leverage their condescension, greed and criminal proclivities against them.

    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  153. APilgrim says:
    @Okechukwu

    The unrelenting & witless ‘racial-IQ-correlation’ denialism of Okechukwu clearly indicates that the fellow is thick & black.
     
    That must be why Nigerian immigrants have an IQ of 110.

    Look, this dog won’t hunt. You don’t even believe it yourself otherwise you would let the facts speak for themselves rather than engaging in constant promotion and propaganda. And you’re not alone either because this idea is a fringe position that almost everyone rejects, including the Africans themselves.

    Africans certainly do not believe that anyone is superior to them intellectually. Often the converse is true. Africans think they’re smarter. It’s this conviction that whites are dumb that drives the yearly multi-billion dollar 419 industry. White marks are called mugus (look it up). I'm not talking about the poorly written letters either. I'm talking about top-level, ingenious, highly sophisticated operations that often take millions of dollars at a time. Many of the victims are the so-called high IQ types: Doctors, bank presidents, oil company executives, politicians, engineers, accountants, etc.

    So the "dumb" Africans are living in mansions and flying around in private jets by virtue of outsmarting and cleaning out the "smart" whites.

    110 is an average IQ for the upper cohort (Bimodal Right Maxima), of American Caucasians.

    A cherry-picked sample of 110 IQ Nigerians represents a serious ‘Brain-Drain’ for Nigeria.

    We don’t need Nigerian O&G. We definitely don’t want any more Nigerians. British slave-traders delivered too many 4 centuries ago. They have only now been ‘bred-up’ to an average IQ of 85.

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  154. APilgrim says:

    About the only sigificant population in Africa, sufficiently intelligent & moral, to merit resettlement in the USA, are the Christian White Boers of South Africa.

    Federal Courts and the Feral Congress will NOT allow them to immigrate. Nor are we allowed to help them there.

    The stupid Black Africans are not worth a dime, nor a damn.

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter Display All Comments
  155. Factorize says:

    Oke, feel the love!

    The latest research has finally allowed us to move beyond the frustrating class and race arguments that have been ongoing for thousands of years. Now that embryo selection has meaningful predictive power, the largely small sub-group difference will soon be seen as not particularly relevant. It seems very unlikely that existing psychometric predictors will be of much use when IQ enhancement of 10-15 points becomes possible. The rational first question that one would ask would be whether one were enhanced or not, not race or class identity. We are now rapidly approaching the time where the main psychometric division will between the enhanced and the unenhanced.

    Once the philanthropy community realizes how great a social investment opportunity genetic enhancement is, a wave of wealth should flow in. With genetic enhancement there would be returns that lasted forever. As soon as people had high homogenous PGS for chromosomes regression to the mean would no longer occur. The rate of return on such investment would be very large. It is not difficult to imagine that a new fundamental human right could soon be enshrined into law: a minimum IQ of 100.

    All of us now need to adjust to the idea that through time, the least intelligent of the enhanced generation will be more intelligent than all existing humans. It would be in all of our self-interests
    to elevate the legal protections of the developmentally disabled, as through time it is almost inevitable that we will all qualify for this label.

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter Display All Comments
  156. Okechukwu says:
    @James Thompson
    Thanks for your comments and links.


    Jelte Wicherts and Richard Lynn had a series of exchanges about African intelligence. You should look at the whole exchange. The link below mentions the final conclusions they reached.

    https://www.unz.com/jthompson/james-watson-nobel-laureate-and-unperson/?highlight=Wicherts

    There is now more up to date material from Rindermann. David Becker is re-doing the Lynn database, and even after cutting out papers which have poor reporting of results, the main findings still stand.

    https://www.unz.com/jthompson/migrant-competence/?highlight=Rindermann+Africa

    If you use the search bar on my blog you will find links to many of the original papers, which is far better than newspaper reports.

    It is unusual to argue that a group of people are bright because they are dishonest, but it is certainly a hypothesis worth testing. In fact dishonesty is linked to lower ability.

    https://www.unz.com/jthompson/do-brighter-minds-incline-to-honesty/?highlight=dishonesty

    Thanks. I will take a look at the links and comment.

    It is unusual to argue that a group of people are bright because they are dishonest, but it is certainly a hypothesis worth testing. In fact dishonesty is linked to lower ability.

    All parties to these transactions are dishonest. The mark’s dishonesty is a critical element in the successful prosecution of the scam. Indeed, their dishonesty is definitely linked to lower ability, such that they are outwitted and destroyed financially.

    Another interesting aspect of 419 is the assumption among the some of the victims that Africans are stupid and corrupt and that crimes against African countries don’t carry any consequences. Imagine plotting with an American functionary to steal tens of millions of dollars from the Federal Reserve. Or plotting with Exxon-Mobil to receive oil under the table at millions of dollars below the market price. These are the kinds of illegal ventures these scam victims are engaged in with respect to Africa. Consequently, the African scam operators are able to leverage their condescension, greed and criminal proclivities against them.

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  157. @APilgrim
    A SNIP Backgrounder:

    Single nucleotide polymorphisms, frequently called SNPs (pronounced “snips”), are the most common type of genetic variation among people. Each SNP represents a difference in a single DNA building block, called a nucleotide. For example, a SNP may replace the nucleotide cytosine (C) with the nucleotide thymine (T) in a certain stretch of DNA.

    SNPs occur normally throughout a person’s DNA. They occur once in every 300 nucleotides on average, which means there are roughly 10 million SNPs in the human genome. Most commonly, these variations are found in the DNA between genes. They can act as biological markers, helping scientists locate genes that are associated with disease. When SNPs occur within a gene or in a regulatory region near a gene, they may play a more direct role in disease by affecting the gene’s function.

    Most SNPs have no effect on health or development. Some of these genetic differences, however, have proven to be very important in the study of human health. Researchers have found SNPs that may help predict an individual’s response to certain drugs, susceptibility to environmental factors such as toxins, and risk of developing particular diseases. SNPs can also be used to track the inheritance of disease genes within families. Future studies will work to identify SNPs associated with complex diseases such as heart disease, diabetes, and cancer.

    APilgrim- thanks for giving a good explanation that most people can follow. I worked in science, excelled in standardized tests in my field.
    I always felt that the best teachers could simplify things so that the person on the street could better understand.
    I have a disdain for those who like to use language and symbols way beyond the expertise of over 90% reading an article such as this one. Most people that use Unz Review are concerned, perceptive individuals. Most are not geniuses nor are they in the superior IQ category, but they have a strong desire to learn why things tick whether in science or in the way societies operate.
    Some people seem to want others to be impressed so they write in a way that few can understand. Perhaps it makes their swollen egos swell a lot more. I bet that most of these people with big egos would be lost if they had to connect pipes that one learns in plumbing 102.
    I am getting off track. Thanks for explaining things so that all types can understand the topic such as SNPs.

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  158. Tim too says:
    @APilgrim
    A SNIP Backgrounder:

    Single nucleotide polymorphisms, frequently called SNPs (pronounced “snips”), are the most common type of genetic variation among people. Each SNP represents a difference in a single DNA building block, called a nucleotide. For example, a SNP may replace the nucleotide cytosine (C) with the nucleotide thymine (T) in a certain stretch of DNA.

    SNPs occur normally throughout a person’s DNA. They occur once in every 300 nucleotides on average, which means there are roughly 10 million SNPs in the human genome. Most commonly, these variations are found in the DNA between genes. They can act as biological markers, helping scientists locate genes that are associated with disease. When SNPs occur within a gene or in a regulatory region near a gene, they may play a more direct role in disease by affecting the gene’s function.

    Most SNPs have no effect on health or development. Some of these genetic differences, however, have proven to be very important in the study of human health. Researchers have found SNPs that may help predict an individual’s response to certain drugs, susceptibility to environmental factors such as toxins, and risk of developing particular diseases. SNPs can also be used to track the inheritance of disease genes within families. Future studies will work to identify SNPs associated with complex diseases such as heart disease, diabetes, and cancer.

    The text is quote from:

    https://ghr.nlm.nih.gov/primer/genomicresearch/snp

    should be attributed.

    Read More
    • Replies: @Lost american
    Appreciate this too. Thank you.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  159. @Tim too
    The text is quote from:
    https://ghr.nlm.nih.gov/primer/genomicresearch/snp

    should be attributed.

    Appreciate this too. Thank you.

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
Current Commenter says:

Leave a Reply - Comments on articles more than two weeks old will be judged much more strictly on quality and tone


 Remember My InformationWhy?
 Email Replies to my Comment
Submitted comments become the property of The Unz Review and may be republished elsewhere at the sole discretion of the latter
Subscribe to This Comment Thread via RSS Subscribe to All James Thompson Comments via RSS