The Unz Review: An Alternative Media Selection
A Collection of Interesting, Important, and Controversial Perspectives Largely Excluded from the American Mainstream Media
 BlogviewJames Thompson Archive
Piffer Rides Again
Predicting group intelligence averages by polygenic risk scores alone.
🔊 Listen RSS
Email This Page to Someone

 Remember My Information



=>

Bookmark Toggle AllToCAdd to LibraryRemove from Library • BShow CommentNext New CommentNext New ReplyRead More
ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
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 three times during any eight hour period.
Ignore Commenter Follow Commenter
Search Text Case Sensitive  Exact Words  Include Comments
List of Bookmarks


The figure shows standardized polygenic scores by population for Education GWAS, in descending order (1000 Genomes Populations, EA MTAG, N= 3,257 SNPs).

One function of a blog is to let people shoot down ideas. Conjectures have a short half-life. Refutations always snap at their heels. David Becker, whose latest version of country IQs received trenchant criticisms, and is now working on all of those (particularly dealing with countries with very low scores), and I will come back to him again when he is ready with the revised edition. Next in the crosshairs is Davide Piffer, who I put up for criticism by one and all in previous years.

http://www.unz.com/jthompson/piffers-equation-further-updated/

Piffer has bounced back with a new paper

https://www.mdpi.com/2624-8611/1/1/5/htm

in which he turns once again to an idea on which he has worked since 2013. He argues that a relatively small number of genetic markers, ranging from 127 to 3527 SNPs depending on the construction method, can be used as stand-ins for a general pattern of selection, which has led to some genetic groups being brighter than others.

Here is the abstract:

Abstract: Genetic variants identified by three large genome-wide association studies (GWAS) of educational attainment (EA) were used to test a polygenic selection model. Weighted and unweighted polygenic scores (PGS) were calculated and compared across populations using data from the 1000 Genomes (n = 26), HGDP-CEPH (n = 52) and gnomAD (n = 8) datasets. The PGS from the largest EA GWAS was highly correlated to two previously published PGSs (r = 0.96–0.97, N = 26). These factors are both highly predictive of average population IQ (r = 0.9, N = 23) and Learning index (r = 0.8, N = 22) and are robust to tests of spatial auto-correlation. Monte Carlo simulations yielded highly significant p values. In the gnomAD samples, the correlation between PGS and IQ was almost perfect (r = 0.98, N = 8), and ANOVA showed significant population differences in allele frequencies with positive effect. Socioeconomic variables slightly improved the prediction accuracy of the model (from 78–80% to 85–89%), but the PGS explained twice as much of the variance in IQ compared to socioeconomic variables. In both 1000 Genomes and gnomAD, there was a weak trend for lower GWAS significance SNPs to be less predictive of population IQ. Additionally, a subset of SNPs were found in the HGDP-CEPH sample (N = 127). The analysis of this sample yielded a positive correlation with latitude and a low negative correlation with distance from East Africa. This study provides robust results after accounting for spatial auto-correlation with Fst distances and random noise via an empirical Monte Carlo simulation using null SNPs.

Piffer explains the use of polygenic risk scores, and says:

The goal of this paper is to test the predictive power of polygenic scores, independently of spatial auto-correlation and noise due to drift and migrations. The prediction is that the polygenic selection model explains average population IQ better than a null model representing only drift and migrations. This implies that the frequencies of alleles with positive effect in the GWAS have different means across different populations.

Piffer accepts that an important limitation of polygenic risk scores based on European DNA is that they may miss other variants. This does not necessarily favour Europeans in comparison with other populations because polygenic scores contain both positive and negative variants. He explains:

For example, a recent GWAS carried out on Peruvians found a population-specific variant that reduces height by about 2.2 cm. Since this variant is polymorphic only in populations of Native American descent, it would have been missed by a European-based GWAS, potentially leading to an overestimation (relative to Europeans) of the PGS for the Peruvian population. A similar scenario might happen with EA polygenic scores, where population-specific variants with negative or positive effects are missed in other populations, leading respectively to over and under-estimations of the non-European population polygenic score. However, since population specific variants can also have a positive effect, the effects will tend to cancel each other out, thus limiting the potential bias.

Evidence suggesting that this is the case can be gathered from the polygenic score on height calculated using an European-based GWAS which produced very low scores for Peruvians, the second lowest in the 1000 genomes samples. Since most GWAS hits are not causal (so-called “tag SNPs”), but are genetically linked with “true” causal variants, and because patterns of Linkage Disequilibrium vary across populations (for example, Africans have on average much smaller LD blocks), this will reduce predictions for populations that are genetically distant from the GWAS sample.

Piffer also used the Lee (2018) data described below:

http://www.unz.com/jthompson/journey-of-1-1-million-miles/

I said then:

At the request of a referee, Lee et al. had 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 disequilibrium) 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 root stock, predict European abilities. All that for later, when better data become available.

In all his calculations Piffer uses height as a control variable, because it is polygenic, has heritable and environmental factors, and can be measured accurately in a non-contentious way. This makes it a good comparison with intelligence, which is also polygenic, also affected by environmental factors and can be measured accurately in a way that some people find absolutely, totally and utterly contentious. See his methods statement in his paper for the construction of the polygenic and socio-economic scales. Piffer uses Monte Carlo simulations to obtain a benchmark for spurious correlations.

Piffer is not trying to predict the IQs of individuals in different genetic groups, but merely their group averages. He is using a simple and restricted set of genetic findings to see if he can predict these averages, arguing that the limited set he is studying are indicators that a much larger sets of genes have gone through an evolutionary process of selections, and are responsible for group differences.

A note on Fst measures: a score of 0 means that two populations are interbreeding freely, which is described as “complete panmixis”. A score of 1 means that the two populations do not share any genetic diversity, and are not interbreeding, as happens with geographic or cultural isolation from each other.

The EDU3 score in Fig 2 is the most reliable. Note that US Blacks and African Caribbeans from Barbados are above the trend line, which Piffer discusses below.

Table 5 and Fig 9 show a fascinating finding: standard psychometrics suggests that the Ashkenazi are the brightest genetic group (estimated IQ 110, which is very close to what the polygenic score predicts (IQ 108). The sample size for Ashkenazis is only 145. The Finnish sample has 1738 subjects. Both are European sub-groups, hence less subject to issues which come from genetic distance (linkage equilibrium decay, population specific variants), but it would be good to have a replication of the Ashkenazi result.

In his discussion, Piffer asserts:

The calculation of population-level polygenic scores (average allele frequencies with positive GWAS beta) is a promising and quick approach to test signals of polygenic adaptation. The results clearly showed population differences in PGS (Figure 3), which correlated with estimates of average population IQ (Figure2) and students performance on standardized tests of mathematics, reading and science (r= 0.9 and 0.8, respectively).

The EDU3 polygenic score is the most robust, and is the best predictor. Monte Carlo simulations strongly suggest that the reported findings are not a fluke. Using this technique to construct a separate equation to predict height succeeds in predicting average population heights. As you might expect, that height equation does not predict average population intelligence, which shows that his equation on intelligence is not simply picking up a simple racial difference, and dressing it up as something which predicts racial differences in intelligence. The mere fact that the equation can predict at least the average intelligence of other racial groups shows that the European origin of the polygenic risk score does not prevent it from having wider relevance. It may well be the case that in all racial groups the genetic variants which boost or reduce intelligence are broadly similar (although they may be located at slightly different points in the genomic sequence).

Piffer identifies a statistical issue with polygenic risk scores: when do you stop? That is to say, how many predictive SNPs should you include? The strongest predictors and no more, or the long list of any that are predictive to any extent? One of his reviewers suggested the following approach.

“Start with the quantile that has the most significant SNPs, and then add quantiles in declining order of genome-wide significance. Initially, adding quantiles will improve prediction, but after a certain point, adding more quantiles will make prediction worse. At that inflection point you have the optimal PGS”.

Piffer found that in his data there was degradation of signal across significance quantiles, as shown by a weak trend for lower significance SNPs to have lower correlation with population IQ. There is flexibility about many SNPs are needed for good predictive power, and this has been discussed in more detail by Steve Hsu, regarding the benefits of compressed sensing.

The correlation coefficients between each score and the population IQ were computed. In turn, the correlation between the correlation coefficient and the quantile was computed, yielding a weak but significant correlation (Spearman’s r = −0.38, p = 0.0152). The PGS generated from most SNP subsets had lower predictive power than that of the full set.

As regards making general racial predictions on the basis of European DNA, the polygenic score is surprisingly good, so much so that one can discuss the findings and the implications. For example, he says:

Indeed, the IQ of African Americans appears to be higher than what is predicted by the PGS (Figure 2), which suggests this cannot be explained by European admixture alone, but it could be the result of enjoying better nutrition or education infrastructure compared to native Africans. Another explanation is heterosis (“hybrid vigor”), that is the increase in fitness observed in hybrid offspring thanks to the reduced expression of homozygous deleterious recessive alleles.

Piffer concludes:

Future GWAS studies should be carried out on non-European populations. Indeed, trans-ethnic GWASs are a promising resource for the identification of alleles with homogeneous and heterogeneous effects and the computation of population-specific polygenic scores. Specifically, they would enable us to include SNPs that are polymorphic only in some populations, and to find the causal SNPs that have the same causal effect in all populations.

In summary:

As Piffer is well aware, many people working in genetic research have not been convinced by his arguments. When he first presented his findings, they identified a number of criticisms, to which he replied. He expected a reply to his comments, but was told to publish a formal paper, which is a reasonable request. Piffer has now had his paper accepted, so it is time to criticize it. Open science, open discussion and may the best arguments win.

 
• Category: Science • Tags: Davide Piffer, Genomics, GWAS, IQ 
Hide 316 CommentsLeave a Comment
Commenters to Ignore...to FollowEndorsed Only
Trim Comments?
    []
  1. dearieme says:

    I hope you had a pleasant Easter break, doc. At least before the news from Sri Lanka.

    Now, a couple of quibbles.

    Africans … our probable ancestral root stock isn’t quite right, is it? We are not descended from present day Africans, but from a small subset of Africans who lived some 50k – 200k years ago. Present day Africans are presumably descended from some of that small subset and from many others of the time.

    Ashkenazis … are European sub-groups My understanding is that Ashkenazis are by descent about half European, half Middle Easterner.

    Lastly a question for your American readers.

    One of Piffer’s reviewers writes At that inflection point you have the optimal PGS. I’ve noticed that many American bloviators refer to inflection points when they mean turning points. When did this habit start, and is it anything more than a silly affectation? Is it, for instance, a sign that the writer is unfamiliar with secondary school calculus?

    • Replies: @James Thompson
    , @res
    , @Anon
    , @utu
  2. @dearieme

    Yes, Easter was fine till Sri Lanka.

    Now, some quibbles with your quibbles.

    I still think “Out of Africa” holds up, though I will watch other hypotheses develop. So I meant root, not current branches. If you want it fancier: Early Africans, before the walkabout.

    Ashkenazis: I may be out of the loop, but I thought they had turned out to be Italians, with little Middle East. Might have just misunderstood a paper from a couple of years ago. Their inter-marriages, even given Jewish restrictions, were with Europeans.

    “Inflection point”. Asymptote. Exponential. I agree that these are cultural appropriations from mathematics. The funny thing is that what the referee meant could be explained in English as “the point of diminishing returns” or “the point when improvements stop and errors increase”. That would be too simple, wouldn’t it?

    • Replies: @res
    , @dearieme
    , @Anon
  3. gcochran says:

    Ashenazi: 60% European, almost all maternal ancestry, most of that Italian. ~40% Middle eastern.

  4. res says:

    One of Piffer’s reviewers writes At that inflection point you have the optimal PGS. I’ve noticed that many American bloviators refer to inflection points when they mean turning points. When did this habit start, and is it anything more than a silly affectation? Is it, for instance, a sign that the writer is unfamiliar with secondary school calculus?

    It looks like the mistaken (non-mathematical) version comes from business jargon: https://www.investopedia.com/terms/i/inflectionpoint.asp

    Yes, it is a sign that the writer is unfamiliar with secondary school calculus (or does not like using precise words). Not sure about what else it might indicate.

    For those who don’t know the mathematical definition, take a look at this page’s explanation of stationary, turning, and inflection points:
    http://bestmaths.net/online/index.php/year-levels/year-12/year-12-topic-list/stationary-and-turning-points/

  5. res says:
    @James Thompson

    I still think “Out of Africa” holds up, though I will watch other hypotheses develop. So I meant root, not current branches. If you want it fancier: Early Africans, before the walkabout.

    This graphic seems like a reasonable way of thinking about it:
    https://en.wikipedia.org/wiki/Human_evolution#Evolution_of_genus_Homo

    https://upload.wikimedia.org/wikipedia/commons/b/b8/Homo_sapiens_lineage.svg

    Caption:

    A model of the phylogeny of H. sapiens during the Middle Paleolithic. The horizontal axis represents geographic location; the vertical axis represents time in thousands of years ago.[136] Homo heidelbergensis is shown as diverging into Neanderthals, Denisovans and H. sapiens. With the expansion of H. sapiens after 200 kya, Neanderthals, Denisovans and unspecified archaic African hominins are shown as again subsumed into the H. sapiens lineage. In addition, admixture events in modern African populations are indicated.

    • Replies: @James Thompson
  6. res says:

    Worth referring to the threads following Factorize’s comment here: http://www.unz.com/jthompson/world-iq-82/#comment-3161778

    Hopefully those discussions will happen here now. I’ll include my response inline here since I am still curious about the answer.

    One thing which surprised me in the Piffer paper is how much lower the height PGS/phenotype correlation was compared to the EAPGS/IQphenotype correlations he sees. Especially given that the EA/IQ connection is a bit indirect.

    The correlation between Height PGS and average phenotypic height was r = 0.486.

    Any thoughts on that? Is there any reason we would expect more population specific SNPs for height than for IQ?

    • Replies: @Merculinus
  7. res says:
    @dearieme

    My comment 4 was a response to this. Sorry about that.

    • Replies: @dearieme
  8. dearieme says:
    @James Thompson

    My understanding is that Ashkenazi descent on the female side is predominantly European – probably Italian or maybe Provencal. On the male side it is predominantly from the Middle Easterners. In the link they reckoned that the Middle Eastern portion was overwhelmingly from the Levant. By this they mean Lebanese, Palestinian, Syrian, and Jordanian, rather than Druze, Egyptian, Bedouin, or Saudi.

    https://journals.plos.org/plosgenetics/article?id=10.1371/journal.pgen.1006644

    • Replies: @RaceRealist88
  9. dearieme says:
    @res

    Nothing to apologise for, old horse.

  10. @res

    Height is more affected by nutrition both across time (huge secular trend) and space (population differences) than IQ.
    Piffer showed that socioeconomic factors explain a lot more of the variance for height than for IQ. Heritability studies also show that there is still significant shared environmental influence on height but not on intelligence

    • Replies: @res
  11. @gcochran

    Thanks. Badly remembered on my part.

  12. @res

    Thanks so much. Finally, something I can put on a T shirt.

    • Replies: @res
    , @Wizard of Oz
  13. res says:
    @James Thompson

    Finally, something I can put on a T shirt.

    Now there is an idea. I wonder how that would go over. Seriously. For example, how would people at most US universities react to that?

    • Replies: @Realist
  14. res says:
    @Merculinus

    Thanks!

    Height is more affected by nutrition both across time (huge secular trend) and space (population differences) than IQ.

    Do you have a quantitative comparison for these? What would be a good measure? Cohen’s D? My sense is the Flynn Effect and corresponding height trend are in the same ballpark. In addition, the between country variation in both seems similar to me. But I have not done a real quantitative comparison myself and would be very interested in being corrected if wrong.

    Piffer showed that socioeconomic factors explain a lot more of the variance for height than for IQ.

    OK. I did not notice Tables 3 and 4 as much as I should have. I wish he had done single variable models using HDI and Total Protein for comparison as well. The full model R^2 were similar for both phenotypes. It is interesting that the Total Protein models ended up with a higher R^2 than the HDI models.

    It would also be interesting to see a scatterplot matrix of the PGS, HDI, and Total Protein values along with each phenotype (2 separate 4 variable scatterplot matrices). A correlation matrix of all of those variables would also be interesting.

    Heritability studies also show that there is still significant shared environmental influence on height but not on intelligence

    Piffer talked about that on page 72. The shared environment for height was about 10% so a good bit less than the difference in variance explained by the two PGS. Though in between population (time or space) comparisons the environmental influence is likely to be larger as you observed.

    That is interesting beyond the context of this paper. Any thoughts on why height and IQ would differ in shared environmental influence? Especially given the greater heritability of height. Perhaps the better question is why the unshared environment contribution is so much larger for IQ? Measurement error?

    P.S. For anyone interested, David Becker’s V1.3.1 spreadsheet has height in 1896 and 1996 for all countries. In the ADDVAR sheet.

  15. Great paper Mr. Piffer.

  16. Realist says:
    @res

    Now there is an idea. I wonder how that would go over. Seriously. For example, how would people at most US universities react to that?

    They would have no clue.

  17. Imagine thinking that IQ tests test intelligence in lieu of construct validity and that genes cause/influence mental abilities when there is no lawlike relation between the mental and the physical.

    “Open science, open discussion and may the best arguments win.”

    Oh, good. Because numerous conceptual arguments refute the claim that the mental is reducible to the physical – i.e., brain states/structure, genes and physiology. So these “gene hunts” for “genes that cause psychological traits” are a fool’s errand. Associations are inevitable – it’s logically impossible for genes to cause/influence mental abilities.

    • Replies: @ThreeCranes
    , @aksavavit
  18. Anon[109] • Disclaimer says:

    I hope no scientists are taking these “result” as serious given serious issue raised with recent GWAS

    • Replies: @Factorize
  19. Factorize says:
    @Anon

    Do you have a URL for that assertion? Are you referring to the July 2018 Nature Genetics article?

  20. @res

    My opinion is that intelligence is sensitive only to malnutrition, but above certain basic requirements, you don’t see additional benefits. That’s why Chinese people from China score as high or higher than Asian Americans, despite having worse nutrition. On the other hand, African Americans score higher than native Africans, even after accounting for their European ancestry, and Asian Americans are taller than their Chinese counterparts.
    Height on the other hand benefits from better nutrition at all levels, so that there is a more linear relationship between protein intake and height. In other words, malnutrition stunts growth, but hyper-nutrition can turn you into a basketball player. Malnutrition can make a child dumb, but hyper-nutrition won’t turn an average kid into Einstein. We see this as an effect of shared environment even in affluent Western countries, where supposedly children who have a diet really rich in protein grow an inch taller than those with a diet that is rich in protein but not super rich, and the Dutch are way taller than the British and the Germans despite being genetically very similar (ok I will admit that genetic differences might explain some of that gap too, but not all).

  21. The Jews belong to the Indo-European subgroup, particularly the Ashkenazi. On the male side their closest relatives are the kurds, who like the persians are indoeuropeans. It is not necessary to assume the khazar hypothesis as Abraham came from near Ur, which is near the kurdish homelands. Even in the beginning they kept their tribe as homogenous as possible. They probably adapted a semitic language in the same way they adapted a german language(yiddish).
    On the female side the Askhenazi Jews are primarily northern Italian, presumably from the time of Rome and the dark ages where they intermarried with local ladies. Until about the 11th century jewishness followed the male line (Abrham begat Isaac who begat Jacob who begat….) and only thereafter it moved to the female line

    • Replies: @dearieme
  22. @Merculinus

    Why has the Flynn effect (intelligence) been so sensitive to the environment (better nutrition, more and better education)?

    Ulric Neisser estimated that using the IQ values of 1997, the average IQ of the United States in 1932, according to the first Stanford–Binet Intelligence Scales standardization sample, was 80. Neisser states that “Hardly any of them would have scored ‘very superior’, but nearly one-quarter would have appeared to be ‘deficient.’” He also wrote that “Test scores are certainly going up all over the world, but whether intelligence itself has risen remains controversial.”

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

    Ashkenazi Jews are believed to have Middle Eastern and European admixtures. How come they are said to have higher average intelligence than Europeans when Middle Easterners are believed to average lower (I do not assert it is not true, but what is the math)?

    Ashkenazi Jews are about 53% European and 47% Middle Eastern by ancestry.

    http://www.unz.com/isteve/time-and-place-of-european-admixture-in-ashkenazi-jewish-history/

    • Replies: @notanon
  23. dearieme says:
    @John Taylor

    On the male side their closest relatives are the kurds No, the male side of Ashkenazis comes from the Levant (Syria, Lebanon, Palestine, Jordan).

    Abraham came from near Ur Abraham didn’t exist. Abraham is from a fairy-story. Moses is from a fairy-story. Joshua is from a fairy-story. Et bloody cetera. There’s no history in Israelite history, as far as anyone can tell, until the Assyrians arrived.

  24. dearieme says:
    @Merculinus

    the Dutch are way taller than the British and the Germans despite being genetically very similar

    This is a recent phenomenon – if you can find a good explanation then scientific prizes should come your way.

    • Replies: @Merculinus
    , @Bonner Tal
  25. @dearieme

    Selection is way faster than most people imagine, and science is quickly coming to understand this.
    This also explains the Ashkenazi advantage.
    Regarding the Dutch phenomenon, I used it as an example of how really high dairy protein intake is better still than medium protein intake for height but not for intelligence

  26. dearieme says:
    @Merculinus

    Is there conclusive evidence that it’s to do with dairy intake? The Scots eat butter on almost everything, or did when I was a boy. They aren’t as tall as the Dutch.

    • Replies: @Merculinus
  27. @dearieme

    Dairy protein, not dairy fat. Butter is fat

    • Replies: @dearieme
  28. dearieme says:
    @Merculinus

    I stand corrected. So, cheese?

  29. @Merculinus

    Dear Merculinus,

    I would have appreciated it if you had replied to me directly. I may have asked you difficult questions to stimulate contemplation on your part.

    But it was not without fruits. We may have to interpret the human genome with caution.

    Yes, the Ashkenazis show a higher average intelligence score compared to Europeans. And Europeans likewise show a higher average score compared to many other races. These have been established.

    But quickly running to “selection” for answers to questions of logic may simply leave us under the illusion of solving problems we lack answers to. We may sometimes need to dig further.

    For example, if “selection” permitted Europeans of average IQ of 80 (the United States was 80 in 1932) to achieve the Industrial Revolution in the 1700s, with its magnitude and scale, why do we not believe that groups with an average score of 80 today can pull a similar feat?
    [I do not want to bring Woodley and Dutton to predict IQs in the 1700s]

    What would we have seen, if we could, if we had peeked into the genome of Europeans in, say, 1932?

    Why have we not experienced another revolution of a magnitude greater than the Industrial Revolution, now that Europeans are 20 points higher on average?

    The Europeans have done very marvelous things that have impacted the life of the common man the world over. Very magnificent things. And we are truly appreciative of the effort. My point is simply that some scientists may need to interpret phenomena with more caution — including genomic predictions.

    Why do the smart fractions (95th percentile) of groups statistically overachieve educationally? Even when their groups have a low mean IQ? Do we not see a driving force acting within these populations and moving the smart fractions to such attainment?

    On the limitation of PGSs based on European DNA that may miss other variants of negative or positive effect, which may have the consequence of favoring Europeans in comparison with other populations, Piffer explained:

    [Since] population specific variants can also have a positive effect, the effects will tend to cancel each other out, thus limiting the potential bias.

    Evidence suggesting that this is the case can be gathered from the polygenic score on height calculated using an European-based GWAS which produced very low scores for Peruvians, the second lowest in the 1000 genomes samples.

    How meticulous is this? Must there always be comparable numbers of variants of negative and positive effects in every group for every trait?

    Why does the average IQ of American Blacks show a deviation from their PGS prediction? Flynn effect to the rescue for Piffer. If phenotype overrides genotype why fixate on the latter and make it a god? Perhaps some of our scientists need more caution and patience.

    My conclusion is that the fuss about average racial intelligence may not be necessary if Europe achieved the Industrial Revolution with an average IQ of 80. We may need more patience to decipher the human genome and its true implications, but before then we will need caution.

    I hope to hear your thoughts as the gentleman you are.

    P.S. Anyone can also join the discussion. Including the experts.

  30. @Chimela Caesar

    Dutton and Woodley actually think that intelligence went down by 10-20 points since the Industrial revolution, so that if IQ was 100, today it is 80. They think that the Flynn effect masks a decline in g, and is only a gain in specific abilities but not general cognitive ability. If they’re right, this would explain why despite a gain of 20 IQ points, our ancestors managed to achieve what they did achieve. In other words, according to their theory, someone with an IQ of 80 born in 1800 would have the same or higher intelligence level as someone with an IQ of 100 born today.
    We can see this paradox at work with the IQ of Nobel laureates. Many of those born in the 1930s, scored around 130 at the time (i.e. Feynmann, Louis Alvarez, William Shockley, E.O.Wilson, James Watson), which would correspond to about 115 in today’s terms, the average for a college graduate. Also a study done on Cambrdige University professors found average IQ of 132 for Physicists and Mathematicians.
    If the person who scores 130 today does so in part because it’s boosted by the Flynn effect-related specific abilities, then their IQ score would a lower reflect an intelligence level than those with the same IQ 50 years earlier, and this can explain why the average IQ test taker who scores 130 today isn’t as smart as those top physicists, or the Flynn-effect corrected counterparts (those who score 115, corresponding to a FE corrected IQ of 130 some decades ago), aren’t as smart as those top physicists either.

    Regarding your question whether the issue of how many of those pop-specific variants are intelligence enhancing or detrimental, this was addressed more meticolously. There appears to be a small bias in favor of beneficial variants, but even after controlling for this, the the Black-White PGS gap is only slightly reduced (2.43% vs 2.32%). See: https://rpubs.com/Daxide/488754

  31. @Merculinus

    In other words, according to their theory, someone with an IQ of 80 born in 1800 would have the same or higher intelligence level as someone with an IQ of 100 born today.

    That is the point!

    Dutton and Woodley were putatively referring specifically to the European groups in their “selection” narative. When asked, psychology researchers say that groups today having relatively low average IQ scores are yet to see Flynn effect run its full course on them. So it follows that these groups are where the Europeans were in the 1800s — meaning similar brainpower disguised in low scores.

    So perhaps current average racial intelligence scores may need to be interpreted with caution, as it follows from the above “axiom” that IQ and g are in flux and loosely coupled in time and space. That is the caution I implore some scientists to heed.

    What then follows? Even if PGSs are claimed to support these average scores, their interpretation with respect to ability to achieve may be subject to flaws.

    • Replies: @Merculinus
  32. @Chimela Caesar

    I think you misunderstood my point. My point is that the European born in 1800 had the same or lower intelligence level as today due to dysgenic, despite their IQ being lower. In other words, according to Woodley, the Flynn effect is void with regards to g. Thus, environment only boosts non-g related components of IQ, but the g part depends completely (bar severe malnourishment or feral children) on the polygenic score plus rare variants.

  33. res says:
    @Merculinus

    Thank you for that link to Piffer’s interesting analysis of Eurasian specific alleles and polygenic scores.

    What struck me most in that writeup is the PGS_difference plot near the end. There is a noticeable group of large difference populations. Almost all of those are Asian, but one is
    IBS Iberian Population in Spain
    I wonder if he is picking up something about Basques (and the overall genetic impact of that ancestral population) in that plot? Any other ideas?

    For the Asians, that looks to me like evidence the PGS may be underestimating them.

    The relatively small difference for African populations is striking given they are even more genetically distant from CEU than the Asians are. Is it possible there is an offsetting environmental effect masking a real difference?

    This also seems like a good place to return to another issue from the earlier thread. In this comment:
    http://www.unz.com/jthompson/world-iq-82/#comment-3167453
    I observed (regarding the ASW population):

    I don’t know how admixed that population was, but I agree with you that a lack of shift in the EDU3 PGS is suggestive. However, this blog post shows a PCA plot indicating significant admixture for the ASW population (more than 25% by my eye, which is more than the 24% average found by https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4289685/ ): http://blog.goldenhelix.com/grudy/admixture-in-reference-populations-1000-genomes-uses-african-americans-in-african-reference-group/

    Not sure how to reconcile those two facts.

    Any thoughts as to how these two facts can be reconciled:
    1. The ASW population is about 25% admixed with whites (25% of the way between the two populations on a PCA plot).
    2. The ASW population PGS scores are almost the same as the African population scores rather than 25% of the way between those and CEU.

    • Replies: @Merculinus
  34. @res

    PGS accuracy declines with genetic distance from Europeans. Notice how ASW has much higher score than other Africans using the unweighted PGS in Piffer’s method (shown in the scatterplot on this blog post), but about the same using the weighted EA_MTAG (shown in the lollipop chart here on this blog post).

    • Replies: @res
  35. res says:
    @Merculinus

    PGS accuracy declines with genetic distance from Europeans.

    That’s the simple theory, but the PGS_difference plot in Piffer’s rpubs doc makes me think we are seeing something more complex. I suspect genetic distance has different implications when the populations are both diverged and in different environments from their ancestral (e.g. Europeans and Asians) than when one population has remained in its ancestral environment (Africans compared to everyone else) since the groups which moved are likely to have different selective pressures than those which remained in the environment they were originally adapted for.

    I think that would lead to the Europe-Asia PGS showing more unique alleles than Africa-(Europe or Asia), Which is what the PGS_difference plot seems to indicate.

    I am not getting the following point.

    Notice how ASW has much higher score than other Africans using the unweighted PGS in Piffer’s method (shown in the scatterplot on this blog post), but about the same using the weighted EA_MTAG (shown in the lollipop chart here on this blog post).

    I see the ASW PGS just above YRI in one plot and between YRI and the rest of Africa in the other (your much higher, I think). But (modulo admixture in ASW) shouldn’t those populations have similar genetic distances from CEU? And if anything, CEU admixture should make the PGS more accurate for ASW than other Africans.

    Any thoughts on my various (here and earlier comment) ASW admixture points?

    P.S. It would be helpful if Piffer could add figure numbers in his rpubs documents. IIRC that used to be a pain for HTML output, but it looks like it has become easy to do recently with the bookdown package: https://stackoverflow.com/questions/37116632/r-markdown-html-number-figures

    • Replies: @Merculinus
  36. @Merculinus

    Let us employ symbols to elucidate our thoughts.

    My point is that the European born in 1800 had the same or lower intelligence level as today due to dysgenic, despite their IQ being lower.

    Let average European IQ and g in the 1800s be iqEu1800 and gEu1800 respectively.

    Let average European IQ and g in the 2000s be iqEu2000 and gEu2000 respectively.

    Let Flynn effect between both periods be e.

    Let dysgenics between both periods be d.

    Apparently, you mean that:

    iqEu1800 = 80

    iqEu2000 = 100

    Loosely speaking, just before the dysgenics:

    iqEu1800 = gEu1800 = 80

    However, today:

    iqEu2000 = (gEu1800 – d) + e = 100

    This implies that:

    gEu2000 = gEu1800 – d

    gEu2000 < gEu1800

    gEu2000 < 80

    And:

    iqEu2000 = gEu2000 + e = 100

    But just for the sake of simplicity, let us say:

    gEu2000 = 80

    So:

    iqEu2000 = 80 + e = 100

    e = 20

    Now let us consider another group today that putatively has not undergone Flynn effect (we are speaking loosely). For the sake of simplicity, let us consider Punjabi, Pakistan (Pp) as a racial group with reported average IQ of 84.

    Dutton and Woodley do not consider Pp to have undergone dysgenics in their theory (loosely speaking also).

    Following the axiom above, we can assert that:

    iqPp2000 = gPp2000 = 84

    But since 84 and 80 are comparable, for simplicity let us assume:

    iqPp2000 = gPp2000 = 80

    In other words, according to Woodley, the Flynn effect is void with regards to g. Thus, environment only boosts non-g related components of IQ

    Following the assertion above:

    iqEu2000 = gEu2000 + e = 100

    and

    iqPp2000 = gPp2000 = 80

    contain equal g’s when e = 20

    So:

    gEu2000 = gPp2000 = 80

    Even though:

    iqEu2000 = 100

    iqPp2000 = 80

    That is the dilemma of the theories. It implies that many IQ scores today may have some air. But also implies that groups that have not undergone Flynn effect such as Punjabi, Pakistan may have comparable innate intelligence (g) as Europeans today despite the groups having different average IQ scores. The Flynn effect also has implications for some Asian groups:

    Similar gains have been observed in many other countries in which IQ testing has long been widely used, including other Western European countries, Japan, and South Korea.

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

    So the current hierarchy of average racial intelligence scores would become suspect with respect to raw brainpower following the above implications based on the various theories.

    but the g part depends completely (bar severe malnourishment or feral children) on the polygenic score plus rare variants.

    The polygenic scores from the Piffer paper validate the current hierarchy of average racial intelligence scores which may appear to be on shaky foundations based on the analysis above with the various theories.

    I have tried not to invent anything in the above analysis. I have only tried to interpret the various theories with symbols. Please point to me anything spurious so that I may make amends.

    If you have any reservations, please reply me with symbols.

    • Replies: @Merculinus
    , @RaceRealist88
  37. @Chimela Caesar

    I get your point. You are assuming that conditions in Pakistan today are the same as in Europe in 1800. Pakistan might not be the most advanced country but they are westernized and they have had access to technology and schools which are probably behind the Flynn effect. So I am not sure that they are FE-free. Also notice that in the scatterplot the PJL and other groups have negative residuals. Their predicted IQ based on the regression is around 88.

    • Replies: @Chimela Caesar
  38. @res

    Look at the second figure in this blog(Scatterplot) which was taken from Piffer’s paper. The US Blacks point (same as ASW) is kind of a positive outlier and this was even discussed in the paper. So it’s not true that their PGS is the same as for other Blacks. That’s the case only for the (weighted) EA_MTAG PGS which Piffer produced specifically for this blog and is not present in his paper. So depending on the choice of SNPs, the ASW score goes up or down. The ASW score is much higher with the unweighted score than with the weighted one. This might have to do with the fact that GWAS Betas lose validity across populations so the unweighted score performs better for non Europeans, whereas the weighted PGS is more accurate for Europeans.

    • Replies: @res
  39. res says:
    @Merculinus

    Look at the second figure in this blog(Scatterplot) which was taken from Piffer’s paper. The US Blacks point (same as ASW) is kind of a positive outlier and this was even discussed in the paper. So it’s not true that their PGS is the same as for other Blacks.

    ASW is an outlier in the y-axis (Population IQ). In the x-axis (EDU3 PGS) it is almost the same as YRI. I don’t see how the outlier (y-axis, population IQ) aspect bears on your PGS point.

    To repeat my earlier statement:

    I see the ASW PGS just above YRI in one plot and between YRI and the rest of Africa in the other

    I am not seeing your “much higher” in that difference. In the lollipop plot above we are talking about an ASW PGS difference of a tenth or two compared to a CEU/ASW difference of about 2.

    I do think the following is possible (but not certain, recall Piffer’s earlier results where the most significant SNPs were better predictors):

    This might have to do with the fact that GWAS Betas lose validity across populations so the unweighted score performs better for non Europeans, whereas the weighted PGS is more accurate for Europeans.

    What I don’t understand is how that affects the relationship between ASW and YRI (etc. in Africa). Unless you are talking about white admixture in ASW which then gets into the other part of my earlier comment.

    • Replies: @Merculinus
  40. @res

    White admixture is 20%, not 25%, for ASW. So they should be 20% of the way between African and CEU scores.The fact they’re lower can be explained by the strong dysgenics that took place among African Americans (recall William Shockley’s figures, but also later estimates, see: https://jaymans.wordpress.com/2012/06/08/dysgenic-fertility-among-blacks-apparently-yes/). At the same time native Africans were living in a more challenging environment so they weren’t subject to the dysgenic trend due to living in an industrialized society.

    • Replies: @res
  41. You admitted earlier the following:

    Thus, environment only boosts non-g related components of IQ, but the g part depends completely (bar severe malnourishment or feral children) on the polygenic score plus rare variants.

    That implies that the Piffer model picks only raw g. And you said about the Piffer model:

    Also notice that in the scatterplot the PJL and other groups have negative residuals. Their predicted IQ based on the regression is around 88.

    How do you reconcile those with your claim below, if Punjabi Pakistan’s supposedly (I mean by you) “Flynn-inflated” average IQ is 84, comparable to the 88 above?:

    Pakistan might not be the most advanced country but they are westernized and they have had access to technology and schools which are probably behind the Flynn effect. So I am not sure that they are FE-free.

    • Replies: @Merculinus
  42. res says:
    @Merculinus

    I took another look at the reference I gave for ASW admixture above: http://blog.goldenhelix.com/grudy/admixture-in-reference-populations-1000-genomes-uses-african-americans-in-african-reference-group/
    and 20% does sound plausible. I was paying too much attention to the arrow rather than the center of mass of the ASW group. Do you have a reference for the 20% though? I would like to have a solid numerical reference rather than eyeballing a plot.

    The dysgenic explanation sounds plausible, but the gap between YRI and CEU in the PGS (lollipop plot above) is 2. So 20% admixture of CEU with YRI should be 0.4 above YRI (above Peruvians). Do you think the dysgenic effect is sufficient to counteract that? Is it possible that the white admixture in ASW is below the CEU average? That might be a partial explanation.

    Another possible wildcard is ASW has some individuals with significant Native American admixture (odd that we don’t seem to see that in the PCA plot, actually, I looked more closely and think I see four of them shifted towards Asian). From the reference I gave above: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4289685/

    Five individuals from the ASW population from the 1000 Genomes Project have poor consistency in their estimates. These individuals have a large amount of Native American ancestry that was not modeled by the 1000 Genomes Project estimates. That these particular individuals were sampled in Oklahoma, and carry significant Native American ancestry, is supported by our own high estimates of Native American ancestry in 23andMe self-reported African Americans from Oklahoma.

    P.S. Is this the appropriate Shockley reference or did you have something else in mind?
    Dysgenics, Geneticity, Raceology: A Challenge to the Intellectual Responsibility of Educators
    https://www.jstor.org/stable/20373194

    This is an interesting excerpt (the 1 point per percent number sounds implausibly high to me for the genetic effect):

    have combined Reed’s findings with Army pre-induction test data in Figure 5 to estimate that, for low I.Q. Negro populations, each 1% of Caucasian ancestry raises average I.Q. by one point.20 I have suggested ways of controlling for the environmental differences to test the reliability of this estimate. An interesting question is the level at which diminishing returns set in; for example, at 40% Caucasian ancestry, would average I.Q. be 110?

    I did not see a numerical estimate for the dysgenic effect (did I miss it?), but there was this:

    If those Negroes with the fewest Caucasian genes are in fact the most prolific and also the least intelligent, then genetic enslavement will be the destiny of their next generation

    • Replies: @Merculinus
  43. Why do Bangladeshi Americans still make more per capita than Mexicans despite the difference? Why are the welfare usage rates, along with crime rates that much lower?

    Also, it would have been nice seeing the numbers for Bengali Indians, to see whether there is a major difference. We may not be a large community, but I M cutiouz to see where I would stand with Gujarati and Bengali Indian genes. Nizari Ismaili Gujaratis tend to be a mercantile community, hence why we practically run Uganda with only 1% of us accounting for about 70% of tax revenue. They were begging for em to return too, poppop rest his soul got a personal invitation if I remeber correctly. Indians in South Africa also tend to come from Gujarati backgrounds, and while making up only 1%, they regardless own 20% of SAs economy.

    Bengalis also don’t do too bad, in terms of the sciences and innovation. Mohammed Yunus is the pioneer of microfinance. Sal Khan has revolutionized online education, and the field of education through KhanAcademy. Jawed Karim is one of the three cofounders of Youtibe, and one of Paypals earliest employees(Paypal mafia). We also have Fazlur Rahman Khan, considered the pioneer of computer aided design in architecture, and considered the Einstein of structural technology and the greatest structural engineer of the 21st century, along with being a major architect. We also have Jagdish Chandra Bose, Satyendra Bose(boson), Amartya Sen(married to a jewess, not impressed), as well as Rafiuddin Ahmed, the guy who estsbilished dentistry and schools of the sort in India.

    Razib Khan, a Bangladeshi Bengali American also found that Bengali Indians also had slightly different phenotype/whatever makeup, with us Bengali Indians with more Caucasoid/Caucasian admixture. Reallu curious where Id stand

    • Replies: @Eagle Eye
    , @Thulean Friend
  44. j2 says:

    It looks like something is wrong with the Ashkenazi IQ in the study, as is typical in these studies.

    The Dunkel et al study used a sample of 53 Jews, but the page
    https://emilkirkegaard.dk/en/?p=7680
    claims that there is an independent verification from the Wisconsin study with 153 Jews and links to
    http://rpubs.com/Jonatan/jewish_pgs
    Looking at the results of Jonathan Pallesen 2018 from the Wisconcinn study the number of Jews was also 53, not 153, and it is not any independent study. The figures of density vs IQ are exactly the same as in Dunkel et al, so this is not any independent verification. Pallesen is the third author in Dunkel et al and the data seems to be the same 53 Jews.

    In Pallesen’s study the Plot of PGS vs IQ entitled Religion and with the axis IQ and PGS do not have a normal distribution of IQ for Christians, therefore Christians comprise several groups. It does not seem to agree with the previous plot with the axes density and IQ, though they should be the same.

    In (4. Additional information) the table of IQs shows that Episcopalian Church has IQ 105, three points below Ashkenazi IQ of 108. As the Episcopalian Church is a US version of the Anglican Church and England is the scale point of IQ 100, this table means that the IQ of Ashkenazi Jews of the USA is 103. Apparently Germans, who moved to the USA, were less talented, more pioneer type.

    Then there is a replication study mentioned in Pallesen with 212 Jews (3. Replication in HRS). The cognition scores differ very little and it points to an IQ difference of 3 points.

    Concerning the polygenic scores, the study in Pallesen divides PGS into two scores, which allows a selection of markers to get a result you want.

    My conclusion is that Dunkel et al is even more suspicious than Piffer ever.

    • Replies: @res
  45. @Chimela Caesar

    It’s partially Flynn-inflated. Flynn effect is a continuous process, it’s not a binary yes or no thing. In some countries it’s still ongoing. What I meant is that when FE will plateau like it did in the West, Pakistan’s IQ will reach 88. Just read Piffer’s paper: he shows that socioeconomic variables (HDI, protein in food, etc.) still account for some of the variance in IQ, so that fills the gap left by the PGS with environmentally-dependent FE gains.

  46. @Merculinus

    What justifies the claim that IQ tests test intelligence in lieu of construct validity?

    https://notpoliticallycorrect.me/2019/04/07/the-lack-of-iq-construct-validity-and-neuroreductionism/

    What justifies the assertion that trait X is “genetic”? What does it mean for a trait to be “genetic”? What’s the a priori justification for privileging genes over any other developmental variables?

  47. @Merculinus

    What do you mean by “selection”?

  48. @Chimela Caesar

    Imagine thinking that mental abilities can be selected when they’re not physical – only physical things can be selected.

    • Replies: @Chimela Caesar
  49. @Merculinus

    ‘g’ is an invention of test construction and item analysis. It’s not a “real” property in the brain.

  50. @Merculinus

    What is ‘g’? Is it a property in the brain? If so, where is it? Are you asserting that the mental reduces to the physical?

    The Flynn effect is due to the rise in the middle class; IQ tests don’t test ability for complex cognition. The IQ testing enterprise that psychologists rely on is a sham.

  51. @Chimela Caesar

    “The polygenic scores from the Piffer paper validate the current hierarchy of average racial intelligence scores which may appear to be on shaky foundations based on the analysis above with the various theories.”

    That they “validate the current hierarchy of average racial intelligence scores” (can’t say they’re “intelligence scores” because IQ tests aren’t construct valid) isn’t evidence that genes cause IQ scores.

    All of this reduces to the claim “Genes cause thinking”, since the main aspect of test-taking is thinking. The claim is patently absurd. Thinking is irreducible to the physical, the mind is not mechanistic. Ross showed on his Immaterial Aspects of Thought that thought isn’t physical—it’s immaterial, and so, genes, physiology, brain states/structure don’t explain IQ scores (what IQ-ists call “intelligence” scores 0even though the tests aren’t construct valid).

    IQ tests are tests of middle class knowledge and skills. Thus, exposure to these types of class-specific items will necessarily be related to test performance. The types of items on the test are class-specific; even so-called “culture-fair” tests like the Raven are biased and employ knowledge structures found in the middle class—indeed, the Raven is the most biased IQ test of them all.

    IQ tests measure distance from the middle class since it tests class-specific knowledge and skills—not “intelligence”. There is even no theory of intelligence or individual differences in it, as admitted by lead IQ-ist Deary in his 2001 introduction to “intelligence.”

    This is further evidence against the claim that IQ tests test intelligence since accepted measures of unseen functions have well-accepted theories behind them. For example the accuracy of thermometers was established without circular reliance on the instrument itself (see Hasok Chang, Inventing Temperature).

    In regard to IQ tests, it is proposed that the tests are valid since they they predict school performance and adult occupation levels, income and wealth. Though, this is circular reasoning and doesn’t establish the claim that IQ tests are valid measures.

    IQ tests rely on other tests to attempt to prove they are valid. Though, as seen with the valid example of thermometers being validated without circular reliance on the instrument itself, IQ tests are said to be valid by claiming that it predicts test scores and life success.

    But due to how the tests are constructed, this is self-fulfilling and does not prove the claim that IQ tests are valid – that they test what they purport to (intelligence).

    So the claim “IQ tests test intelligence” cannot be true because they are not construct valid.

    P1 If the claim “IQ tests test intelligence” is true, then IQ tests must be construct valid.
    P2 IQ tests are not construct valid.
    C Therefore, the claim “IQ tests test intelligence” is false. (modus tollens, P1, P2)

    • Replies: @Chimela Caesar
  52. j2 says:

    I looked at Dunkel et all, especially what was the PGS they used. It is not the one Piffer used and therefore their result should not be drawn to the same plot as Piffer’s results. It is very interesting that they did not use any known PGS for educational attainment but made their own. The information is said to be in:
    https://www.ssc.wisc.edu/wlsresearch/documentation/GWAS/Lee_et_al_(2018)_PGS_WLS.pdf
    but you look in vain for the exact SNPs that were selected and for the weights. Furthermore, their evaluation criteria allow subjective evaluation, self-estimated
    Lee et al.1 conducted genome-wide association analyses of four phenotypes: educational attainment (EduYears, N = 1,131,881), cognitive performance (CP, N = 257,841), self-reported
    math ability (MA, N = 564,698), and highest-level math class taken (HM, N = 430,445).

    See what you think.

  53. @gcochran

    Remind us Greg. Is that based on the most recent DNA based studies? Can you please give some guidance as to best evidence, including understanding it, even though I have been content on UR threads to cite you, from memory, as authority enough. Actually memory tells me one of your posts had 58 per cent where you now say c. 60.

  54. @res

    Would it be surprising to find that natural selection favoured genes and epigenetics which made sure lack of a good diet had less of an effect on brain growth and function than on physical size of the rest of the body?

    • Replies: @res
  55. @James Thompson

    May I Dr T, as one of your regular and sympathetic readers who is not in the daily habit of using the jargon, acronyms, abbreviations and symbols familiar to you and your graduate students ask that, if only to be more powerfully influential, you present UR readers with something which doesn’t just read like your notes preparatory to providing a peer review? A regularly attached extended glossary might be an efficient and economical way for you to help your readers. When I say “extended glossary” I mean one which might not only spell things out but give illustrative calculations and/or hypothetical counter-examples to better define meaning.

    • Replies: @James Thompson
  56. https://www.telegraph.co.uk/news/science/science-news/12061787/Intelligence-genes-discovered-by-scientists.html

    https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4270739/

    The fail here is in uniqueness. No study has located a unique genetic paring, dynamic or formation that is unique to any particular group of humans. That might be genetic. But to date none such exists. The most obvious differences that turn out IQ standouts seems to be smaller closed or close knit communities that are mated and environmentally con-formative, especially in training and education.
    ——————————————–

    I was intrigued by the discussion regarding score variances in relation to any society. i.e. Pakistan. Assuming exposure to either industrialization, western thought should be reflected across the board is to assume that either of these conditional variables exist as components across the entire population. Pakistan remains a highly class structured society as does India, Japan, Vietnam, etc.

    A factor such as literacy rate is going to play a major role in any test scenario.
    __________________________

    • Replies: @Anon
  57. @res

    I recently read that ASW White admixture is 20% but I can’t recall the reference. I am sure you can find it if you google it.
    And yes assortative mating for intelligence or education suggests that Whites that admixed with Blacks were lower than the White average.
    Regarding Shockley, I was referring to the charts Shockley shows in the video interview where he shows least educated Black women having the most children. This can be found on Youtube at this link: https://www.youtube.com/watch?v=sAszZr3SkEs#t=2m5s

    • Replies: @res
  58. @Merculinus

    You admitted earlier the following:

    Thus, environment only boosts non-g related components of IQ, but the g part depends completely (bar severe malnourishment or feral children) on the polygenic score plus rare variants.

    That implies that the Piffer model picks only raw g. And you said about the Piffer model:

    Also notice that in the scatterplot the PJL and other groups have negative residuals. Their predicted IQ based on the regression is around 88.

    How do you reconcile those with your claim below, if Punjabi Pakistan’s supposedly (I mean by you) “Flynn-inflated” average IQ is 84, comparable to the 88 above?:

    Pakistan might not be the most advanced country but they are westernized and they have had access to technology and schools which are probably behind the Flynn effect. So I am not sure that they are FE-free.

    • Replies: @Chimela Caesar
  59. @RaceRealist88

    Well, it may be “logically impossible for genes to cause/influence mental abilities” but it’s not physically so.

    • Replies: @RaceRealist88
  60. @ThreeCranes

    There is no lawlike relation between the mental and the physical so genes can’t cause/influence mental abilities.

    • Replies: @annamaria
  61. @Chimela Caesar

    Please sorry to everyone. I was behind the conversation on a stale tab so I reposted a comment already responded to.

  62. @RaceRealist88

    I understand how you feel. I think that, generally, it is proper to employ ratiocination to weigh all claims, including those based on the “selection” narrative.

  63. annamaria says:
    @RaceRealist88

    “There is no lawlike relation between the mental and the physical so genes can’t cause/influence mental abilities.”
    — RaceRealist88, have you ever saw a child with Down syndrome or Williams syndrome?
    You are obviously on a wrong forum and in need to find people that share your system of beliefs.

    • Replies: @annamaria
    , @RaceRealist88
  64. @Merculinus

    I have read Piffer’s paper.

    [Piffer] shows that socioeconomic variables (HDI, protein in food, etc.) still account for some of the variance in IQ, so that fills the gap left by the PGS with environmentally-dependent FE gains.

    This would only hold in situations where a population is underperforming compared to their true genetic potential due to, say, severe prenatal malnutrition. But it should not include IQ score gains due to, say, more and better education.

    Let me know if my assertion above is valid.

  65. res says:
    @Merculinus

    Any thoughts on how evenly the Flynn Effect applies across the IQ spectrum?

    For example, this from Wikipedia: https://en.wikipedia.org/wiki/Flynn_effect#Rise_in_IQ

    Some studies have found the gains of the Flynn effect to be particularly concentrated at the lower end of the distribution. Teasdale and Owen (1989), for example, found the effect primarily reduced the number of low-end scores, resulting in an increased number of moderately high scores, with no increase in very high scores.[15] In another study, two large samples of Spanish children were assessed with a 30-year gap. Comparison of the IQ distributions indicated that the mean IQ scores on the test had increased by 9.7 points (the Flynn effect), the gains were concentrated in the lower half of the distribution and negligible in the top half, and the gains gradually decreased as the IQ of the individuals increased.[16] Some studies have found a reverse Flynn effect with declining scores for those with high IQ.[17][13]

  66. res says:
    @Wizard of Oz

    Not at all surprising. That would be my working hypothesis ; ) But I am curious as to how true it is.

  67. res says:
    @Merculinus

    Thanks. I went looking for ASW admixture references.

    This one: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3392282/
    Says:

    Figure 4 shows the results for the four HapMap populations ASW, CEU, CHB and YRI using markers on chromosome 1, as an example of two-way admixture. Here, CEU and YRI were used to represent the European and African ancestries, respectively, of the ASW population. CHB was included in the analysis to provide an additional axis of variation. Most of the ASW samples were found to be restricted on the cline of CEU and YRI; the distance between the centroid of ASW and the line segment connecting the centroids of CEU and YRI is 0:002, much smaller than the length of the line segment (0:11). The results confirmed that the ASW population is a genetically recent admixture with an average of 19:2% of European and 80:8% of African ancestry, as calculated using Equations (18) and (19) with P3 representing the centroid of ASW. The individual admixture proportions were calculated also using Equations (18) and (19) and are listed in Table S1. The results of STRUCTURE were also shown in Table S1. The average admixture proportions estimated by STRUCTURE were 26:3% and 73:7%, for European and African ancestry, respectively. The Spearman correlation coefficient between the results of PCA and those of STRUCTURE was R~0:847.

    I also downloaded their Table S1 which gives individual admixtures from ASW. They average 24.2%

    Not sure which of those numbers should be considered most accurate. Figure 4 is using only Chromosome 1 (not sure about the other analyses).

    Thanks for the YouTube Shockley link. The number of children difference is stark, but I did not see a good way to estimate the quantitative effect on IQ (qualitatively it is obvious).

  68. Anon[308] • Disclaimer says:
    @dearieme

    Africans … our probable ancestral root stock isn’t quite right, is it? We are not descended from present day Africans, but from a small subset of Africans who lived some 50k – 200k years ago. Present day Africans are presumably descended from some of that small subset and from many others of the time.

    Now, some quibbles with your quibbles.

    I still think “Out of Africa” holds up, though I will watch other hypotheses develop. So I meant root, not current branches. If you want it fancier: Early Africans, before the walkabout.

    The out of Africa theory has not held.

    *Strictly toward an effort to simplify my below commentary that prefaces the later cited research, “hominid” refers to any homo species that is not modern human / homo sapiens sapiens (ie: includes Neanderthal, etc). I also regularly use the term “race” to simplify what could otherwise be qualified by better, specific, but technical terms.

    It seems to me that the sum of the below cited research points to a process of hominid-human cross breeding as the mechanism primarily responsible for any significant genetic differentiation across races.

    This does not discount an evolutionary process that occurred somewhere, as the below cited research seems to be uncertain as to of the origins of “cro magnon” (this group’s early historical form, as possibly implied by the overall picture painted by the below cited research, may be best categorized as the modern human group with the least hominid admixture; ie: the unmixed humans species).

    Again, the below cited research seems to combine for a foundation for a reasonable hypothesis that modern human racial differentiation is not due to evolution but instead due to inter-species cross breeding and the resultant admixture quality and quantity.

    Admixture that will have been competitively reduced to varying common percentages in geographically isolated groups (eg: the research states that African archaic hominid admixture can reach percentages that are 2-4 times the highest noted Neanderthal admixture percentage in European groups. Specific data would be needed to determine such percentages for any group, the admixture percentages cannot be assumed).

    Research conclusion: Europe was the birthplace of mankind, not Africa.

    https://www.telegraph.co.uk/science/2017/05/22/europe-birthplace-mankind-not-africa-scientists-find/

    https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0177127

    Research conclusion: No known hominin is ancestor of Neanderthals and modern humans (Cro Magnid).

    https://phys.org/news/2013-10-hominin-ancestor-neanderthals-modern-humans.html

    Research conclusion: “Caucasians” did not evolve from Africans and likely did not migrate out of Africa, but into Africa.

    http://www.scirp.org/Journal/PaperInformation.aspx?paperID=19566

    A territorial origin of haplogroups α- and β-remains unknown; however, the most likely origin for each of them is a vast triangle stretched from Central Europe in the west through the Russian Plain to the east and to Levant to the south. Haplogroup B is descended from β-haplogroup (and not from haplogroup A, from which it is very distant, and separated by as much as 123,000 years of “lat- eral” mutational evolution) likely migrated to Africa after 46,000 ybp. The finding that the Europeoid haplogroups did not descend from “African” haplogroups A or B is supported by the fact that bearers of the Europeoid haplogroups, as well as all non-African haplogroups do not carry either SNPs M91, P97, M31, P82, M23, M114, P262, M32, M59, P289, P291, P102, M13, M171, M118 (haplogroup A and its subclades SNPs) or M60, M181, P90 (haplogroup B), as it was shown recently in “Walk through Y” FTDNA Project (the reference is incorporated therein) on several hundred people from various haplogroups.

    Research conclusion: Genetic evidence for archaic admixture in Africa

    http://www.pnas.org/content/108/37/15123.full

    Research conclusion: Early modern humans mated with Homo species in Africa. Sub-Saharan genetic diversity is due to introgression with archaic Hominids.

    https://www.biorxiv.org/content/early/2018/03/21/285734

    Research conclusion: 28,000 yr old European Cro Magnid genetics are unchanged from modern European genetics.

    https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0002700

    The Paglicci 23 individual carried a mtDNA sequence that is still common in Europe, and which radically differs from those of the almost contemporary Neandertals, demonstrating a genealogical continuity across 28,000 years, from Cro-Magnoid to modern Europeans.

    Research conclusions (from the conclusions cited at Wikipedia): Neanderthal-derived and Denisovan-derived ancestry is significantly absent from most modern populations in Sub-Saharan Africa.

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

    The observed excess of genetic similarity is best explained by recent gene flow from Neanderthals to modern humans after the migration out of Africa.

    The findings show that the source of modern human gene flow into Neanderthals originated from a population of early modern humans from about 100,000 years ago, predating the out-of-Africa migration of the modern human ancestors of present-day non-Africans

    My summary: European and Asian mixing with Eurasian hominids (Neanderthal and Denisovan) did not occur in Africa (as evidenced by the fact that Africans largely do not carry these genetics), but at the same time would have had to occur well before the proposed Out of Africa date.

    • Replies: @RaceRealist88
  69. aksavavit says:
    @RaceRealist88

    “it’s logically impossible for genes to cause/influence mental abilities”

    Is the brain size of chimps not determined by genes, and aren’t chimps mental abilities determined by (inter alia) their brain size?

    • Replies: @Anon
    , @RaceRealist88
  70. Anon[421] • Disclaimer says:
    @James Thompson

    James
    Trying to help unravel the Ashkenazi ancestry puzzle which describes my own DNA.
    I have read much about it and my favorite is “The Invention of the Jewish People”
    by Prof. Shlomo Sand. The key point is that the European Jews were descended from
    a diaspora people who were living throughout the Roman and Greek cities even before
    the Christian era. Their middle eastern DNA was already mixed with southern
    Europeans even before being invited to migrate north into what became Germany by
    Charlemagne in 850. The word “Ashkenaz” refers to the North Country, not north of
    Judah but north of the Roman Empire. Jewish men took North European wives, which
    would explain a study indicating a large measure of Flemish ancestry in their MitDNA.

  71. @Wizard of Oz

    Thanks for your comments. I had been having similar thoughts last night, while preparing a further post on Piffer, in which an intelligence research takes a sympathetic but critical view of the paper, and then Piffer replies.

    The problem is that their discussion is pretty technical. It has to be, because that is where the real issues lie. I can provide a glossary, but haven’t done so because really it should be done by a geneticist. A good and simple explanation may be available somewhere, just not known to me. I had one go at it here:

    http://www.unz.com/jthompson/genetics-made-very-simple/

    I also mentioned the problem here: http://www.unz.com/jthompson/more-genes-for-intelligence-a-pattern-emerges/

    For those who, like me, don’t take easily to genetic jargon, just think of all this as computer code. If you look through traditional computer code you will find sub-routines and Go To instructions. Some of the code is embedded in sub-routines, some code acts as signposting, and some code does the essential processing work along the way. All these sections of code can develop a flavour: by the time you get to a distant sub-routine your variable names will have drifted further down the alphabet; line numbers will be higher, the types of calculations will have altered, and will involve the products of far previous sub-routines. Of course, the genetic code is nothing like this, but linkage can be close or far, common or unusual, and if a piece of code gets picked out because it is particularly useful, it can carry some neighbouring useless code with it, like fluff on a toffee. New ways of understanding and analyzing the code are being developed fast, so new findings may well arise when analysis moves from exploratory association work to laboratory manipulations of individual genes in Petri dishes.

    When I make simplified explanations I check them with the author, but I am aware that I should do more. A few days ago I had started, and then stopped working on, an explanation of linkage disequilibrium. Right now, I would like to start from scratch and say “Variations at different positions in the genetic code can be independent and random; or paired together in a linked and synchronous manner”. Linkage disequilibrium refers to the second case, but the phrase sounds more like it should be the first.

    In truth, my explanations would be quickly rejected by geneticists.

    So, I agree with what you are saying, but don’t have a solution at the moment.

  72. Anon[429] • Disclaimer says:
    @aksavavit

    Brain size, architecture, and thus fundamental intelligence and behaviors are determined by genes across all species.

    No cat has the genetically conferred ability to learn arithmetic, and none ever will.
    No human has the genetically conferred ability to see the far infrared spectrum, and none ever will.

    Neither are all human brains physically the same. Those physical differences limit the operating window.

    Genetically conferred intelligence and instinct (animal behavior) can be environmentally shaped within a certain window that is strictly limited by the physical brain, and more so in terms of the self-aware species (human) control over their choice of behaviors, but not outside of that window.

    These two comments by “RaceRealist 88” make him not worthy to reply to. You don’t need to debate such flawed views of what even constitutes a valid proof. They are self-refuting:

    Because numerous conceptual arguments refute the claim that the mental is reducible to the physical

    it’s logically impossible for genes to cause/influence mental abilities.

    • Replies: @RaceRealist88
  73. @annamaria

    Those are genetic disorders not psychological traits/mental abilities.

    “You are obviously on a wrong forum and in need to find people that share your system of beliefs.”

    “I just want an echo chamber where everyone believes what I do.”

    What are my beliefs?

    • Replies: @annamaria
  74. @Anon

    Please. More than a few teeth are needed for that claim to have any force. In any case it’s false.

    https://notpoliticallycorrect.me/2017/10/22/more-than-a-few-teeth-are-needed-to-rewrite-human-evolutionary-history/

    • Replies: @Anon
  75. @aksavavit

    Genes don’t “determine” anything. Are you a genetic determinist?

    You’re talking about cognition (an intentional state). Genes partly build brains that have the ability to engage in intentional states (human brains). Intentional states are irreducible.

    • Replies: @Anon
  76. @Anon

    There are no ex anti specifiable ceteris paribus clauses with respect to mental phenomena so there are no psychophysical laws. Since there are no psychophysical laws, therefore, mental traits cannot be selected.

    Genes don’t “determine” anything. Reductionism (including psychophysical reductionism) is false.

    • Replies: @Anon
    , @Anon
  77. utu says:
    @dearieme

    I’ve noticed that many American bloviators refer to inflection points

    You expect too much from the reviewers of obscure journal such as ‘psych’ that so far has published total of 10 papers out of which three have Emil O. Kierkegard as author/co-author and two are by Richard Lynn. Only one paper comes from a respectable author which is James Flynn who actually demolishes Rushton in his paper:

    https://www.mdpi.com/2624-8611/1/1/3/htm
    To assess his contribution I argue: (1) That the racial ranking for desirable traits is not as tidy as it seems; (2) That the ice ages hypothesis has been falsified; (3) That the black/white Q gap is more likely to be environmental, with black American subculture as the culprit; and (4) That appeals to correlated vectors and regression cannot disentangle genetic and environmental causes.

    All 1o papers so far accumulated a total of 3 citations of which one goes to Flynn and two go to Piffer of which one was by an un-affiliated individual called Miller in Mankind Quarterly that as wiki has it

    https://en.wikipedia.org/wiki/Mankind_Quarterly
    [Mankind Quarterly] has been described as a “cornerstone of the scientific racism establishment” and a “white supremacist journal”,[1] “scientific racism’s keepers of the flame”,[2] a journal with a “racist orientation” and an “infamous racist journal”,[3] and a “journal of ‘scientific racism’”.

    • Replies: @James Thompson
  78. j2 says:

    About Piffer’s method.
    If you look at the plot Plot of PGS vs IQ in
    http://rpubs.com/Jonatan/jewish_pgs
    it is clear that for individual the correlation of PGS vs IQ is very weak. It this plot it seems that the used PGS gets high values on Ashkenazi Jews, but it does not correlate with IQ, i.e., the data points for Jews spread to the right direction, but not up.

    Piffer’s method seems to be like this: as educational achievement scores give high values for East Asians and low for Sub Saharan Africans, the PGS used includes SNPs that correlate to ethnic composition and therefore also to educational achievement. For this reason the average values for different nations plot nicely on a straight line, while individual data points are quite much spread.

    As Finns have some East Asian admixture, this method raises Finnish PGS, even thought the reasons for good PISA results may be quite unrelated to it.

    In the earlier PGS used by Piffer British scored very much the same as Finns in PGS. Now, however, the PGS has been modified in order to include Ashkenazi Jews. British score now lower than Finns. What apparently has been done is that as Ashkenazi Jews have Middle Eastern admixture, the new PGS includes SNPs that reflect this admixture. Piffer carefully chooses not to measure any Middle Eastern people with this new PGS, as they would also go up in educational achievement. In this way, omitting the possible counter examples from the Middle East, the new PGS places Ashkenazi Jews where the researchers want to put them, even though the Plot of PGS vs IQ looks like it simply reflects genes that Ashkenazi Jews have more commonly than British.

    So, this method is matching the PGS to get the result you want, it is so all the way. It is so also in Piffer’s paper showing that height correlates slightly negatively with IQ. The explanation is naturally that there are three main races (black, white, yellow) and for white IQ correlates positively with IQ, while the two other races need a different line as they are different genes.

    This is pseudoscience, this time I agree with debunkers.

    • Replies: @utu
  79. notanon says:
    @res

    Any thoughts on why height and IQ would differ in shared environmental influence?

    height as a function of many potential nutritional factors vs IQ as a function of the presence or absence of a handful of very specific nutrients?

    so height more analogue, IQ more digital?

    • Replies: @res
  80. notanon says:
    @Chimela Caesar

    Why has the Flynn effect (intelligence) been so sensitive to the environment (better nutrition, more and better education)?

    presence or absence of iodine?

    since WW2 there’s been a big increase in fish consumption in some regions, plus iodized salt in others.

    Ashkenazi Jews are believed to have Middle Eastern and European admixtures. How come they are said to have higher average intelligence than Europeans when Middle Easterners are believed to average lower?

    my understanding is the way merchant-bankers operated in this era was a merchant would give gold to a banker in city A in exchange for a note which he could exchange with a banker in city B for gold and generally the two bankers were from the same family (as that was the only people you could trust).

    so what if the Jewish-Italian thing was a business merger?

    i.e. the sons of Jewish merchant-bankers from the Levant marrying the daughters of Italian merchant-bankers in Italy to create a new link in the trade chain?

    if so maybe both sides of the pairing were +1SD above their respective base populations.

    • Replies: @Chimela Caesar
    , @j2
  81. David says:

    Here’s something interesting by my standards. The OED only has two words that start with “pif.” One, piffle and the other pift. The first is senseless or stupid talk and the second the sound an arrow makes in flight.

    To me this suggests that the sound “pif” is not one English speakers like.

    I hope Mr Piffer will not be offended by the observation.

  82. notanon says:
    @Chimela Caesar

    My conclusion is that the fuss about average racial intelligence may not be necessary if Europe achieved the Industrial Revolution with an average IQ of 80.

    if a population has a standard curve with an average of 80 then the higher IQ outliers would follow one pattern

    but if a population had a more bimodal distribution where the right side of the curve was “normal” for an average IQ of 100 but the left side was distorted lower through missing a critical nutrient then the actual average might be lower even though the distribution of high IQ outliers was the same as if the average was 100.

    • Replies: @Chimela Caesar
  83. @RaceRealist88

    The polygenic scores from the Piffer paper validate the current hierarchy of average racial intelligence scores which may appear to be on shaky foundations based on the analysis above with the various theories.

    My use of “validate” there had the sense of “trying to make valid”. Thanks

  84. @notanon

    Thanks for the thoughtful responses. Many would rather resort to an easy “selection” narrative.

  85. Anon[703] • Disclaimer says:
    @RaceRealist88

    Double down and crank your philosophical bullshit posing as science as much as you like, ad nauseam.

    You can even graduate from 10, now to 20, and later to 30 cent words and phrases in the effort. It makes no difference to your essential lack of content. Your masturbation doesn’t make up for statements that don’t meet the standard of horse sense let alone the processes, terms, and historical results of recorded observation.

  86. @notanon

    The size of a population’s smart fraction matters.

    if a population has a standard curve with an average of 80 then the higher IQ outliers would follow one pattern

    The size of the smart fraction would remain the same.

    but if a population had a more bimodal distribution where the right side of the curve was “normal” for an average IQ of 100 but the left side was distorted lower through missing a critical nutrient then the actual average might be lower even though the distribution of high IQ outliers was the same as if the average was 100.

    The size of the smart fraction would be halved. The loss is due to damage to potential candidates on the left side.

    So monetary or technological economy of the population in the first case would be twice that of the second owing to size. Roughly speaking.

    • Replies: @res
    , @notanon
  87. Bonner Tal says: • Website
    @dearieme

    I kinda doubt the Dutch are taller than northern Germans. There is a notable North-South cline of height in Germany. (And a South-North cline in IQ.)

    • Replies: @res
  88. Anon[312] • Disclaimer says:
    @RaceRealist88

    You follow your philosophy, which has no material (scientific) basis and is only a non-observable (invented) concept, with a conclusion employing the misuse of the now loaded term “reductionism” in a context that is wholly political (I’m sure you’d argue philosophical – which in the context of science is about the same) and not related to the standards of observation (science).

    In other words, your mental toys followed by a straw man is not a successful argument. Try departing for a forum where your Latin thesaurus wins arguments by virtue of its presence and in spite of a lack of content. You’ll be happier.

  89. Anon[505] • Disclaimer says:
    @RaceRealist88

    Genes don’t “determine” anything. Are you a genetic determinist?

    There’s the Stalinist style interview posing as rhetoric.

    To feel entitled to pose this question, which functions as a moral test, in an ostensible debate means that someone has spent too much time swimming the the modern political murk in either internet forums or academia. Its in his eyes and perspective has been lost. The dead body of science is rotting at the bottom of the pool.

    These type of questions would have you disrespected and ignored in any serious world, unless you were merely too young to know better and made a mistake. An intentional test would mark you as a non-scientist.

    Intentional states are irreducible.

    Ignoring your misuse of words in a science argument (like irreducible) once again:

    Which you require philosophy to “prove”. And therefore your asserted proofs are scientifically irrelevant.

    Which means that your assertion would be laughed out of any room of serious scientists.

    You are trying to train philosophy on human free will in an effort to assert new axioms of biology.

    That effort marks you as a politician or a pure philosopher who prefers notion to data. People that science was designed to exclude.

  90. Anon[586] • Disclaimer says:
    @RaceRealist88

    “More than a few teeth” is not any formal science standard of which I am aware. Its a political statement, or at least a statement that is worded to have a political effect, the likes of which we are continuously remind us that you are prone and attracted to.

    The only difference is that author attempts to use unorthodox folksy language to bolster the impact of his dismissal, where you attempt to bulldoze your way through arguments using unnecessarily florid language even for academic discussion (a high bar). The aim is the same: an avoidance of the science-standard clarity that would reveal the arguments for what they are.

    Second, there are “more than a few teeth” within the other research that I cited and you ignored.

    While I don’t have the practical space here and the readers don’t have the attention (few do) for a full research review, my briefer reference list was enough for you to know better than to try to dismiss the entire post, citing no less than six studies, with the word “please”, a repetition of the lame phrasing of someone you googled who is critiquing one study, and an unsupported statement.

    You are bad at this.

    Do you know how to read and critique research? Do you know what the standards are? I get that you are attempting to put on an air like you do. However underneath the curious confidence that likes to express itself with unsupported assertions that no serious science trained person would make, I am detecting that you do not and that you are not trained at all.

    Its okay. People try to bullshit their way through an understanding of science here all of the time whenever discussions over politically contentious science occurs. Its not unusual. Just know that it comes off like a non-mechanic trying to bullshit their way through an engine disassembly.

    Last, the cited politically correct “evolutionary history” as is shaky (my word) theory (science’s word). Its not verified factual history.

    Again, your loose use of language and that of those whom you reference does not help your case. The other studies that I cited, which you ignored, makes my case for my opinion. Try that format some time: hypothesis, foundation, conclusion.

    Do so and you will have made your first baby step as a scientist, away from being a mere political commentator on science (ie: noise in the signal), and away from being mocked by anyone with an education.

  91. res says:
    @notanon

    Interesting. I had thought about that nutritional difference before, but never connected it with shared environmental influence. Thanks.

    What seems counterintuitive to me is that if this is true then IQ should be easier to improve (over poor environment baseline) if we know the specific micronutrients (Omega 3 fatty acids seem like an obvious possibility, any other ideas?).

    • Replies: @notanon
  92. res says:
    @Chimela Caesar

    The size of the smart fraction would be halved. The loss is due to damage to potential candidates on the left side.

    Depends on how much good genes/good environment are correlated. In a heavily class based society I think that correlation is likely to be high. Which would leave the smart function relatively unimpaired. This would also explain why the Flynn Effect would be concentrated at the lower end of IQ.

    • Replies: @Chimela Caesar
  93. res says:
    @Bonner Tal

    There is a notable North-South cline of height in Germany. (And a South-North cline in IQ.)

    Is there any good research/data covering this? Seeing IQ and height varying in opposite directions in fairly closely related populations is interesting.

  94. Anon[217] • Disclaimer says:
    @EliteCommInc.

    Pakistan remains a highly class structured society as does India, Japan, Vietnam, etc.

    Japan has among the highest wealth/HDI equality in the developed world. By what metric is it even remotely comparable to India and Pakistan in terms of class structure?

  95. notanon says:
    @Chimela Caesar

    well i can’t claim to fully understand the stats but seems to me if

    1) a population has a fixed genetic potential

    but

    2) kids need a sufficient amount of nutrient x to reach their full potential

    then three options like
    – the whole population has sufficient x
    – the whole population is deficient
    – the wealthy segment has sufficient but the rest are deficient
    would give three different distributions?

    • Replies: @Chimela Caesar
  96. And we are importing Indians by the millions to help us “innovate” in the tech sector, and to be our doctors.

    Where does that put our real IQ?

    • Replies: @Wizard of Oz
  97. notanon says:
    @res

    iodine has been known about for a long time

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

    (hence the possibility that increased fish consumption post WW2 in some regions is possibly a factor in the Flynn effect)

  98. annamaria says:
    @RaceRealist88

    “Those are genetic disorders not psychological traits/mental abilities.”

    — But according to your expressed beliefs (see you post # 64), “There is no lawlike relation between the mental and the physical so genes can’t cause/influence mental abilities.”

    You need to figure out how “genetic disorders” (the physical phenomenon) are not related to the mental abilities of individuals with Down syndrome and Williams syndrome. Or you may simply continue believing into your model of un-relatedness between the physical (which you have identified as “genes”) and the “mental” (including intellectual abilities).

    PS: “In biology, a gene is a sequence of nucleotides in DNA or RNA that codes for a molecule that has a function. Genes can acquire mutations in their sequence, leading to different variants, known as alleles, in the population. These alleles encode slightly different versions of a protein, which cause different phenotypical traits.” https://en.wikipedia.org/wiki/Gene

    “Down syndrome (DS or DNS), also known as trisomy 21, is a genetic disorder caused by the presence of all or part of a third copy of chromosome 21. It is typically associated with physical growth delays, mild to moderate intellectual disability, and characteristic facial features. The average IQ of a young adult with Down syndrome is 50, equivalent to the mental ability of an 8- or 9-year-old child, but this can vary wide.” https://en.wikipedia.org/wiki/Down_syndrome

  99. utu says:
    @j2

    This is pseudoscience, this time I agree with debunkers.

    I am glad that you see it.

  100. Eagle Eye says:
    @BengaliCanadianDude

    One more Bose: Amar Bose, MIT prof and founder of Bose Corporation, “born in Philadelphia, Pennsylvania,[3] to a Bengali Hindu father, Noni Gopal Bose and an American mother of French and German ancestry, Charlotte. ”

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

    • Replies: @BengaliCanadianDude
  101. @BengaliCanadianDude

    Why do Bangladeshi Americans still make more per capita than Mexicans despite the difference?

    Because you’re not reading the chart properly. It shows Bengali Bangladeshis in their native environments vs Mexicans in LA. Bangladeshi-Americans are the elite of their countries, which is why you’d expect them to make more money than mexicans in the US (who largely come from poorer strata).

    A lot of poorly functioning countries send their elites to Western countries (Pakistan and India are two other examples). That says nothing about the coginitive level of their respective populations at home, which is what you’d really want to measure if you want an accurate understanding of their capabilities. Bangladesh has almost 170 million in population. If their native IQ was similar to NW Europeans or East Asians they would have been an economic superpower by now.

  102. @res

    I agree with you in principle. If a society has high good genes/good environment positive correlation, its members would “sort” themselves along the distribution into a normal curve.

    But for the scenario that was pictured:

    a population had a more bimodal distribution where the right side of the curve was “normal” for an average IQ of 100 but the left side was distorted lower through missing a critical nutrient then the actual average might be lower even though the distribution of high IQ outliers was the same as if the average was 100.

    It appears to be a case of lack of “opportunity” with its attending consequences. The size of the smart fraction would be reduced.

    • Replies: @res
  103. @A citizen over 21

    Assuming they average IQs of 112 from gene pools of average 106 – not unreasonable assumptions I am sure you would agree – then that should put American average IQ up for at least a couple of generations.

  104. j2 says:
    @notanon

    “i.e. the sons of Jewish merchant-bankers from the Levant marrying the daughters of Italian merchant-bankers in Italy to create a new link in the trade chain?

    if so maybe both sides of the pairing were +1SD above their respective base populations.”

    The Italian admixture in Ashkenazi Jews is dated to 600-800 AD. That means, before there were Italian city states with merchant bankers. The most likely origin of Ashkenazi Jews is that they migrated from the Bysantine Empire to Italy, which was at that time under a Germanic tribe Langobards. But Ashkenazi Jews could not marry Langobards as Langobards had just converted to Catholicism from Arian Christianity. Therefore Jews living among Langobards could only marry their own or make children with slave girls. The italian mixture probably comes from Italian slave girls, that is, from Italians (Romans) that Langobards had very recently conquered. Because of this origin, Ashkenazi Jews did not start with 1SD above 100. They more likely started a bit under 100 because of Middle Eastern admixture.

    As for the theory that Ashkenazi Jewish intelligence would have increased because of marriage selection and the rich having more children, persecutions killing the more stupid, or rare diseases having a side effect of raising the IQ of heterozygotes, one can easily discard every one of these explanations.

    Firstly, every Jewish man was expected to marry and to have children. Kahal provided social support for poor families. Ashkenazi Jewish population grew 1.5% yearly, much faster than other European populations. It is very clear from this that the selective pressure for higher IQ in the Ashkenazi Jewish population was smaller than in other ethnic groups. There was less competition for survival among the Ashkenazi.

    After Ashkenazi Jews mover to Poland their position was very good, they were protected by the king and treated as half nobles because of their money. There was only one major persecution, the Khmelnyctsky 1648 uprising were some 20,000 Jews died (some 1/10 of 200,000 Polish-Lithuanian Jews), but that was followed by the Swedish Deluge (the war) where about 1/3 of the whole Polish population died. Therefore, that time did not put any special selective pressure on the Jews. The second persecution that caused a major loss of life was the Holocaust. Most European nations have faced worse catastrophes and have died more in wars than Jews, who were exempted from military draft.

    The theory that Jewish diseases have a side effect of increasing IQ for hererozygotes (heterozygote advantage) is shown incorrect by measurements of the carrier frequency and the disease prevalence. If there were heterozygote advantage, the ratio of these two would show it, but the measured ratio is the same as without heterozygote advantage.

    The only explanation that remains is that the American Ashkenazi Jewish higher IQ is a result of selective migration to the USA from Russia in the end of the 19th and beginning of the 20th century. Jews in Russia had had guaranteed employment by the state. When this privilege, and some other privileges, like the alcohol monopol, were removed, many intelligent Jews found themselves without employment and migrated to the USA. In the case of Ashkenazi Jews the more intelligent ones migrated to the USA, while typically the ones who migrate are more enterprising but not the most intelligent as the intelligent fraction has good positions in the home country. This selective migration fully explains the measured higher IQ of the US Ashkenazi Jews, which probably is today around 103-104, the same as with Ashkenazi Jews in Israel, but earlier was higher, around 110.

    This rubbish with Ashkenazi Jewish IQ is just a typical myth to explain why the take over of science and other fields is not a take over but a result of meritocracy.

    • Replies: @res
  105. @Thulean Friend

    I’d argue that the people coming from more wealthier backgrounds in these countries tend to represent the genetic IQs of their respective populations, as these people tend to have the money, the care, the avvailability of food, access to clean water, actual schools and the likes of things. Of course some tribal girl living on a dollar a day, nomadic lifestyle, drinking from disease-ridden streams, going to low ranked, non accredited, weak educational institutions run by NGOs and volunteers won’t have the same IQ as some other girl living in a clean neighbourhood in Kozhikode, with the pumps running, sewage in good working order, as well as being connected to the electricity grid, while having the opportunity to attend those famous “global schools” we hear so much about. Im obviously not referring to the top 1%, but those of the upper middle income class as well as those part of the middle middle income class. Obviously not Korean IQ or anything, but nothing this low.

  106. @Eagle Eye

    I have Bose headphones and a speaker as well. I also regularly use Paypal to…pay, and I have KA on my phone for the sake of education. And I have Youtube. Lovely…….

  107. @notanon

    then three options like
    – the whole population has sufficient x
    – the whole population is deficient
    – the wealthy segment has sufficient but the rest are deficient
    would give three different distributions?

    Yes. The first two are of one type — a normal distribution. The third is of another — a bimodal distribution, theoretically. But its curve would be shaped by the effect size of x deficiency.

    But the first two normal distributions would typically (meaning highly likely) vary in the values of the measures of their central tendency like the mean, median and mode due to the effect of x deficiency.

    • Replies: @notanon
  108. Anon[150] • Disclaimer says:

    You can clearly see that South Asians are not part of the trendline. The South Asians have higher educational attainment than “IQ” would predict. In fact, we are just below Europeans. It is also very curious that he left out South Asians from various analyses. I bet he did not like the results.
    Besides that, how much of the variation is explained by GWAS? They have not hit even 10% yet. Further, the correlation between education and IQ is 0.65 from what I recall.
    You also need to study populations separately. The African prediction will not work if you are not going to look at the SNP’s that are unique to them.
    Just because the guy managed to publish a paper does not mean that it is good. This one is crap, especially because South Asians are not part of the trend. It is obvious to even a blind squirrel.

    Also, it seems pretty obvious that genes are doing a poorer job of predicting South Asian educational attainment. You simply know this from the fact that South Asians in the US are maybe the most educated group out there. So the fact that the prediction is not accurately measuring their educational attainment is problematic.
    Basically, what you are seeing is that polygenic scores have less predictive power when it comes to South Asians. I look here mainly at Gujaratis from Houston. The vast majority of them probably have a degree of some kind.

  109. Anon[150] • Disclaimer says:

    I just took a quick look at Lee’s paper. The analysis was restricted to European populations only. Somehow this dumbass pfeiffer decided to make a prediction on non-Europeans based on European data. If anything, I am now convinced that South Asians will get the same score as Europeans if you account for novel genes that are unique to South Asians.

    • Replies: @notanon
    , @James Thompson
  110. notanon says:
    @Chimela Caesar

    But the first two normal distributions would typically (meaning highly likely) vary in the values of the measures of their central tendency like the mean, median and mode due to the effect of x deficiency.

    yes – i was thinking there might be some way of proving the effect of an x factor (if it exists) in the mathematical “shape” of the various curves.

    • Replies: @Chimela Caesar
  111. notanon says:
    @Anon

    what you are seeing is that polygenic scores have less predictive power when it comes to South Asians

    caste effects?

    might polygenic analysis of a caste society be similar to analysing different races in that each caste might have different mutations?

  112. notanon says:
    @Anon

    Somehow this dumbass pfeiffer decided to make a prediction on non-Europeans based on European data.

    you mean the study implies there are fundamental genetic differences between populations?

    golly.

    • Replies: @South Asian Dude
  113. res says:
    @Chimela Caesar

    We are reading the excerpt you quoted differently. I read it as the right side (> 100) of the distribution is exactly the same as a normal distribution with mean 100. That would leave the smart fraction unchanged. It is the left side (< 100) that is distorted.

    Whether or not that is an accurate model is another question.

    • Replies: @notanon
    , @Chimela Caesar
  114. @utu

    Thanks for the link to James Flynn’s paper.

  115. @Anon

    Lee also uses the European prediction to estimate the intelligence of an African sample. He can predict 11% of European intelligence, and 1.6% of African intelligence. There is general agreement it would be good to have DNA and educational attainment figures (and IQ measures) on 1 million Africans and 1 million Chinese. One could then test predictive equations within and between continental genetic groups.

  116. @Anon

    South Asians weren’t left out of the analysis, you little conspiracy theorist/nationalistic brat. Simply they weren’t part of some datasets. 1000 Genomes includes them, gnomAD doesn’t. Go and email the Broad Institute and complain that they were racist to leave out South Asians from their samples!
    And Gujarati do have higher polygenic scores than other groups. They simply aren’t as high as East Asians or Europeans.

    • Replies: @South Asian Dude
  117. Anonymous[182] • Disclaimer says:
    @James Thompson

    Yes, but the reduction depends on PGS construction techniques. With causal or highly significant SNPs, reduction in accuracy is much less, that is, the accuracy for Blacks is much higher than one fifth (I’ve seen the results from people who are working on this). Piffer used both causal and significant SNPs, hence the results are trans-racially valid.

  118. notanon says:
    @res

    that was what i meant, yes

    (or some variation on that theme e.g. regional differences in the x factor rather than class differences)

  119. @James Thompson

    Yes, however, Lee et. al clearly mention in their paper that the cohort was exclusively European. Furthermore, if it is predicting only 1.6% of African intelligence, then it really cannot be that predictive.
    The issue is that to properly study African IQ (or for that matter South Asian or East Asian) you need to first have the data of educational attainment. Once you have this data, then you are able to figure out the alleles responsible for intelligence. With a million samples, you barely get a little more than 10% predictive power. So how much data do you need? This study was able to identify a few thousand snp’s. It appears we are talking about alleles in the neighborhood of 10000 or more to get higher predictive power. What complicates this prediction is environment. There are some areas that are more malnutrioned than others. How are you going to factor in this big problem? Now, if you can use a group like Gujaratis from Houston, or the Telegus from the UK, that will actually control for environmental differences. However, here the problem is that South Asians in general are way more educated than the general population.
    Based on the fact that Indians in the United States are highly educated (more so than any other group, you can clearly see that the regression is not making a good prediction.
    This study is bad.

    • Replies: @James Thompson
    , @res
  120. @Merculinus

    For a Northern European supremacist, you really lack some critical thinking skills.. How are you going to claim that the study is good if it is not making a good prediction on Indians living in the United States? Do you realize how educated Indians in the United States are? You resort to ad hominem attack because you don’t have the answer. For the study to be accurate, Indians in the US need to be on the top. This is a study based on educational attainment.

    • Replies: @Merculinus
  121. @notanon

    No, the study is not properly accounting for genetic differences. Don’t you get it? There are genetic differences between Europeans and non-Europeans, and the study is making prediction on non-Europeans by using European genetic data. There are alleles out there that are unique to Africans, South Asians, Europeans, etc. You need to account for those differences. The fact that Piffer (or Pfeiffer or Feiffer, whatever his name is) is not accounting for these differences makes this paper proplematic.

    • Replies: @notanon
    , @Merculinus
  122. @Anonymous

    Thanks. Do you know when these results might be published?

  123. @Anonymous

    You say that there are highly significant SNPs for which the accuracy for blacks is much higher. Uhhhh, how the fuck do you know what SNPs are accurate for Africans if they have not been studied?
    It is simple, really. If you want to study Africans accurately, you need to study a million inidividuals of African descent. There may well be alleles unique to Africans that have higher predictive power.
    Can you prove to any of us here that they don’t exist? How do you know that?

    • Replies: @Merculinus
  124. @South Asian Dude

    Yes, environments still vary considerably, mostly between Sub-Saharan Africa and the rest of the world, though on many health and education indicators those gaps appear to be closing. (There is a new paper suggesting that higher school participation in poorer countries is not leading to a commensurate gain in skills, but I have yet to get into the detail).

    As to how many subjects are required for predictive power, current estimates are 1+ million. However, improvements in analysis (Steve Hsu has outlined the basics of compressed sensing) may mean that much greater accuracy can be achieved for intelligence even at 1.1 million, given that we can already predict European’s heights to within about an inch. People working in the field think that variance accounted for will increase considerably (almost double, some say) and we will see if that proves to be the case over the next few years.

    • Replies: @South Asian Dude
  125. notanon says:
    @South Asian Dude

    i agree with that.

    Somehow this dumbass pfeiffer decided to make a prediction on non-Europeans based on European data.

    i’m saying testing this prediction and proving it wrong isn’t necessarily dumbass.

    • Replies: @South Asian Dude
  126. @South Asian Dude

    Indians in the US need to be on top because you’re Indian or South Asian or whatever? How ridiculously ethnocentric!

    • Agree: res
    • Replies: @South Asian Dude
  127. @South Asian Dude

    Some colleagues have gotten access to an African sample and validated the Lee et al. SNPs on them. Simple.

    • Replies: @res
  128. @South Asian Dude

    Haven’t you seen this publication by Piffer? He computed polygenic scores after removing SNPs that are non African-specific (that is, absent from Africans). This creates a level ground between Africans and non Africans, and the White-Black gap was only slightly reduced (from 2.43 to 2.32%)!
    Link: https://rpubs.com/Daxide/488754

    • Replies: @South Asian Dude
  129. @James Thompson

    Yes, the accuracy will improve with lower sample count as we learn more about these alleles. I can see that. We are kind of in agreement I think that groups need to be studied separately to identify these unique alleles.
    If you go to dna.land, they have a test to predict IQ based on I think 16 alleles. That really is not going to be accurate. We are now thinking 1000’s of SNP’s. So what can you find out from less than 50 alleles? Nothing. It is entertaining, but that’s it. It will not predict your actual IQ.
    I am seeing articles out there that are suggesting that people will start using genetics to predict a child’s IQ by testing them in schools. I don’t see that happening at all. It’s going to take decades. Maybe it will be more accurate for Europeans, but who else is being studied?

    • Replies: @res
  130. res says:
    @j2

    The theory that Jewish diseases have a side effect of increasing IQ for hererozygotes (heterozygote advantage) is shown incorrect by measurements of the carrier frequency and the disease prevalence. If there were heterozygote advantage, the ratio of these two would show it, but the measured ratio is the same as without heterozygote advantage.

    Could you expand on this? Are there any research papers discussing it?

    • Replies: @j2
  131. res says:
    @South Asian Dude

    if it is predicting only 1.6% of African intelligence, then it really cannot be that predictive.

    It is important to realize there is a difference between being able to predict individuals accurately (1.6%) and being able to predict differences between group means. Piffer is doing the latter. Statistical averaging across a large sample can be powerful.

    So how much data do you need?

    This is an interesting question. Steve Hsu et al. are actively looking at this in their compressed sensing work. It turns out the UKBB was enough for height, but not for EA (Educational Attainment). I’m curious whether IQ will be easier or harder than EA (what is the status of large sample GWAS with a decent IQ proxy other than EA?). In Western population EA pretty much turns into a variable with only a few bins.

    • Replies: @South Asian Dude
  132. @notanon

    The problem is, he tried to correlate gwas prediction with population IQ. Basically, he wanted to show that these alleles can predict ethnic IQ. The population IQ itself is pseudoscience with a lot of questionable studies being used. Looking at some of the studies, what you will notice is that a lot of studies in developing countries were conducted in rural areas. To top it off, a lot of studies are about nutrional deficiencies.
    If you are going to compare populations, at least these pseudoscientists should try to be truthful. For developed countries, I counted just one study in a rural area, and it was in Alaska–so I assume Eskimos (who incidentally have a very big head size). None of the European samples were from rural areas.
    Looking at the data, there is a clear difference in IQ between rural and urban areas. I noticed a difference of around 10 in developing countries.

  133. res says:
    @South Asian Dude

    but who else is being studied?

    There is a simple solution. Have Indians, Africans, etc. do studies on their own populations. I really hate the Catch 22 of Europeans being racist if they do studies on non-Europeans and also being racist if they don’t.

    P.S. Do you really not understand the difference between predicting the IQs of individuals and Piffer’s work?

    • Replies: @South Asian Dude
  134. @Merculinus

    No, they need to be on top because of educational attainment. Isn’t the purpose of this study to predict educational attainment based on alleles? Explain to me how the study is accurate.

    • Replies: @Merculinus
  135. res says:
    @Merculinus

    I would be interested in any more details you can share. And please comment here if/when they publish.

  136. @Merculinus

    No, it does not create a level playing field. Just because these alleles have higher predictive power in Europeans doesn’t mean it has higher predictive power in Africans. So, simply speaking, what allele has higher predictive power in Africans? Just because they removed non-African specific alleles means nothing. The point is that African may have alleles out there that have much higher predictive power. How do you know they don’t have alleles with higher predictive power that are unique to them? These alleles only are significant to European populations. You don’t know their predictive power in non-Europeans.

    • Replies: @Merculinus
  137. @res

    I know the difference between individual prediction vs group prediction. Do you understand that for group prediction Lee et Al used in the upwards of 1 million samples? Why did they use just European samples? Because they are studying a group. In the Sciences, you don’t study people individually. Do you think they studied the APOE gene (Alzheimer’s-related gene) based on one individual?

    • Replies: @res
  138. @res

    Oh, they can conduct studies on Europeans, but don’t try to make predictions on non-european data by using European data. That’s called stupidity.

    • Replies: @res
  139. @South Asian Dude

    Causal variants identified by the GWAS have the same effect across ethnicities, and Piffer showed these have a larger Black and White gap. Full stop.

    • Replies: @South Asian Dude
  140. @South Asian Dude

    I am sorry your pet group’s PGS doesn’t score at the top. Even white supremacist can accept that East Asians outsocore Whites, but you can’t accept that your group is outscored.

    • Replies: @South Asian Dude
  141. j2 says:
    @res

    “Could you expand on this? Are there any research papers discussing it?”

    I calculated it, you can calculate it if you want, or check my calculations (but they are correct), it is simple math, see
    http://vixra.org/abs/1812.0166
    Being on pension I do not care to publish papers in journals, especially as editors in journals do not even pass to referees papers that show any errors in main stream myths, like not agreeing with the sky high IQ of Ashkenazi Jews or a smaller than accepted death toll of the Holocaust, or anything that may be against the official myths.

    • Replies: @res
  142. @Merculinus

    “Causal variants identified by the GWAS have the same effect across ethnicities, and Piffer showed these have a larger Black and White gap. Full stop.”

    Oh, so he showed that they have a large black and white gap. But did he show that these variants have the same effect across ethnicities? Or are you pulling that out of your ass?
    You cannot back the above statement, period. You really don’t have a response to my argument, which is that you do not what effects these variants have on non-Europeans. If you believe in genetic differences, then you need to believe that Africans (or South Asians, or East Asians) have variants for intelligence that will not be found in Europeans. You cannot measure the effect size of these variants on non-Europeans by studying Europeans. Only idiots believe that.
    Even the author of the OP has said that Africans need to be studied (a count of a million to get a good picture). Here you are acting like some sort of an expert.
    The cohort is European-only. PERIOD. You don’t have a good argument.

  143. @res

    Res, we are thinking along same lines.

    Say, the sum total t of the population is under two “bells”, the distorted first on the left, and the normal second on the right.

    Let the total smart fraction of t, ceteris paribus, be s. (That is, assuming the whole population was normal.)

    Let the smart fraction sizes under both bells respectively, ceteris paribus, be s1 and s2. Such that s = s1 + s2. (That is, assuming the whole population was normal.)

    Now considering the original population with a distorted left bell and a normal right bell:

    s2 remains unchanged.

    s1 shrinks because of impairment.

    Such that (s1 + s2) < s.

    In the above state, we can say that the size of the smart fraction has been reduced.

    • Replies: @res
  144. @Merculinus

    Do you understand that the study is predicting educational attainment based on variants? Given that Gujarati Indians from Houston will by nature have higher educational attainments than basically every single ethnicity (including the Chinese or Japanese), it is not predicting their educational attainment very well. Do you understand that, you idiot? Here, read this:

    https://www.usinpac.com/indian-americans/census-2010/education-levels/

    That is the education level of Indian Americans. The genetic data predicts them as having lower educational attainment than an average White Joe. Uhh, that is a bad prediction.

    • Replies: @res
  145. res says:
    @South Asian Dude

    I know the difference between individual prediction vs group prediction.

    Your comment illustrates that you do not.

    Do you understand that for group prediction Lee et Al used in the upwards of 1 million samples?

    They were doing prediction of individuals within a group. To make that comparable to Piffer’s work one would look at how well the average prediction across the entire group matched the average phenotype of the group. I’ll bet that match would be rather good (would make an interesting experiment). Though should probably do it on an out of sample group of the same race.

    Why did they use just European samples? Because they are studying a group.

    No. They did it because population structure can affect GWAS results. Using a racially uniform population avoids this.

    In the Sciences, you don’t study people individually. Do you think they studied the APOE gene (Alzheimer’s-related gene) based on one individual?

    That you give this silly strawman makes clear you really don’t understand what I am saying.

    Let’s try again. Simpler this time.

    Predict individuals – Generate a PGS value for an individual then compare it to that individual’s phenotype. This is what Lee et al. did for many individuals within a single group.

    Predict groups – Generate a PGS average value for a group then compare that to the group’s average phenotype. This is what Piffer did for multiple countries.

    Some additional notes.

    – The group prediction I describe can be done either by averaging individuals, as I believe Piffer did, or by plugging group average allele frequencies into the predictor. It would be interesting to see some of those latter numbers added to Piffers’s plots for reference. The ability to use allele frequencies is a nice feature of additive genetic PGS.

    – Hopefully it is obvious that in the group prediction case there is a statistical averaging effect which should reduce noise and increase accuracy. If it is not obvious I would recommend taking a statistics course.

    – One wrinkle in Piffer’s work is the per country genetic and phenotype groups are not identical (different samples from the same countries). This is a potential source of error. Though the previously mentioned statistical averaging should help with that if both samples are reasonably representative (a nontrivial if).

    P.S. The arrogance of moderately skilled South Asians is something to behold.

    • Replies: @South Asian Dude
  146. res says:
    @South Asian Dude

    That is the education level of Indian Americans. The genetic data predicts them as having lower educational attainment than an average White Joe. Uhh, that is a bad prediction.

    You are assuming Indian Americans have the same average genetic scores as Indians in India. That is a big assumption given genetic stratification in castes and the differences in caste numbers between India and the US. Not to mention individual differences in those who choose to come to the US.

    You are familiar with the idea of selective immigration, right?

    • Replies: @South Asian Dude
  147. res says:
    @South Asian Dude

    Oh, they can conduct studies on Europeans, but don’t try to make predictions on non-european data by using European data. That’s called stupidity.

    You (and most of the other critics here) completely miss what is most interesting about Piffer’s work.

    That he gets such good results.

    Perhaps try explaining why he sees such large correlations given the issues raised by the critics. That is the interesting question.

    • Replies: @South Asian Dude
  148. res says:
    @j2

    Thanks for the link. I’ll try to take a look at the paper in the next few days.

  149. @notanon

    I guess one would need an experiment where there are more than one groups of same racial ancestry. One group would need to be the control group assumed to be free of the x deficiency while the others are not. Measurements taken could then be used to plot the curves or analyze the x deficiency effect.

  150. res says:
    @Chimela Caesar

    The smart fraction would only be under the unchanged right half (two halves of one normal distribution, not two separate normal distributions) of the bell (e.g. the top 50% contains the top 5%). No reduction.

    • Replies: @Chimela Caesar
  151. As regards finding a paper which explains GWAS procedures and terminology (Table 1), and which also lists all the problems about drawing conclusions, giving a cautious and skeptical treatment of polygenic risk scores, this has been suggested:

    https://academic.oup.com/emph/article/2019/1/26/5262222

  152. @res

    They were doing prediction of individuals within a group. To make that comparable to Piffer’s work one would look at how well the average prediction across the entire group matched the average phenotype of the group. I’ll bet that match would be rather good (would make an interesting experiment). Though should probably do it on an out of sample group of the same race.

    No, they weren’t doing a prediction of individuals within a group. They took a sample of a million individuals along with their educational attainment to figure out what variants are responsible for educational attainment. So by nature, the cohort of a million is a training set. You can only make a prediction AFTER you have figured out the weights. So, now if you take this trained weights and try to figure out the educational attainment of Gujarati Indians, THAT is a prediction. And the prediction is wrong because the data is based on Europeans. Do you understand now? The way p-value works is the larger the sample, the lower the number. So the cutoff they use tends to be a very low number (to the power of -8 from what I recall). You look at something called a mahattan plot to look at the outliers.

    P

    Predict groups – Generate a PGS average value for a group then compare that to the group’s average phenotype. This is what Piffer did for multiple countries.

    This makes ZERO sense if said group has not been studied before. The effect size is not going to be the same as Europeans. Do you understand that?

    Hopefully it is obvious that in the group prediction case there is a statistical averaging effect which should reduce noise and increase accuracy. If it is not obvious I would recommend taking a statistics course.

    Which again makes zero sense if said group has not been studied before.
    Let me give you a machine learning example. if I train my data for different types of cars, what will the result be if I use a chair as a test? You get garbage results.
    It’s simple really, garbage in, garbage out. That is the Piffer result.

    P.S. The arrogance of moderately skilled South Asians is something to behold.

    No arrogance here. The bottomline is, garbage in, garbage out. The OP agrees that other groups need to be studied due to unique variants. You don’t seem to grasp this concept. Not my problem.

    • Replies: @res
  153. @res

    You are assuming Indian Americans have the same average genetic scores as Indians in India. That is a big assumption given genetic stratification in castes and the differences in caste numbers between India and the US. Not to mention individual differences in those who choose to come to the US.

    They don’t. You can easily see that because Pakistanis are from Pakistan and Bengalis are from Bangladesh. On the other hand, Gujaratis and Telegus are based outside their respective countries. They score higher on educational attainment. However, the problem is when comparing Gujaratis from Houston to let’s see a European group. Here the prediction is horrible given that Gujaratis from TX will have a very high educational attainment.

    • Replies: @Merculinus
    , @res
  154. @res

    Perhaps try explaining why he sees such large correlations given the issues raised by the critics. That is the interesting question.

    It is not a good correlation. Look at Sri Lankans. They are way outside the trend line. Also, I don’t see Gujaratis and Telegus as that far away. If you include SNPs unique to these groups, I pretty much guarantee that the difference from Europeans will disappear.

    • Replies: @res
  155. @South Asian Dude

    Yet you have got to admit that the Gujarati PGS is higher than that of other South Asians. The polygenic score in Lee et al.’s paper (which I am dead sure from your lack of understanding that you’ve not bothered to read) also predicts intelligence, besides educational attainment. We can actually say educational attainment is a proxy for intelligence in this case. I am sorry dude if your pet group doesn’t top the charts. Now, what is the Gujarati’s IQ? Probably lower than 100.
    Lee et al. didn’t use 1.1 million individuals as a training set. They used a sub-group as training set and tested the results on other samples. You are totally ignorant both of theory and methods behind Lee’s work, and Piffer’s, and even basic statistics. Your nickname says blatantly that your attachment to your ethnic group is clouding your reasoning.
    Piffer has already provided a level playing ground between Africans and Europeans by removing from the PGS the variants that are absent from Africans, in this way the PGS is composed only of variants shared by Africans and Europeans. The correlation between the PGS and IQ is still 0.88, and the Black-White gap marginally reduced. How is this not a level playing ground? You’d have to argue that Africans have more population specific intelligence enhancing variants than Europeans do, because if they have the same number as Europeans, the PGS computed from the shared SNPs is unbiased. I don’t see why Africans should have more beneficial mutations, which is the shaky ground on which your attack on Piffer stands.
    You also have to explain how come Ashkenazi Jews have such a high PGS and high IQ, or Finnish people.
    And yes, the causal variants have been shown to explain a substantial amount of variation also among African individuals. I am privy to an unpublished study that explains this.
    If you can’t explain why Piffer’s PGS are so damn good (correlation 0.9) at predicting IQ across different datasets, you can just shut up. And the results aren’t even post-hoc because they are almost identical to work that was done in 2015 by Piffer.

    • Replies: @South Asian Dude
  156. res says:
    @South Asian Dude

    OK. Focusing on Gujaratis from Houston your point makes more sense. So this population:
    http://www.internationalgenome.org/data-portal/population/GIH

    Do you know that group in particular has a high educational attainment? Given the overall Indian American EA stats I would expect so, but it would be good to know how typical GIH (and the larger local group it was sampled from) is compared to that.

    It is worth noting Piffer left GIH off his Figure 2 IQ/EDU3 scatterplot. Presumably because of the lack of accurate phenotypic data.

    Regarding your overall point for GIH in particular, I would agree the GIH EDU3 average looks a bit low. Hard to say more without more information about that specific community.

    One of the problems with EA as a metric rather than IQ is that EA has a significant (much more so than IQ IMHO) cultural component. Presumably the EA GWAS is picking up some combination of IQ/conscientiousness/overall health/etc. SNPs. But how those SNPs translate into EA will vary greatly depending on how much the surrounding culture values and encourages education.

    • Replies: @Merculinus
  157. res says:
    @South Asian Dude

    Do you know what “correlation” is in a statistical sense? Outliers don’t disprove a large numerical correlation (though they might make one wonder what is going on there).

    It seems you are mostly concerned about the unexpectedly low Indian results. There is a simple solution for that. Have people do studies on Indians (both in India and the US if possible).

    P.S. I would guess the Sri Lankan scatterplot outlier is driven by exactly what I mentioned before–selective immigration. I’m assuming the IQ data was from Sri Lanka (should probably check the paper, but too busy right now). The population itself is STU Sri Lankan Tamil in the UK

    Just checked. From the paper:

    The Sri Lankan UK population also constitute an outlier, because their IQ is lower than that predicted by the PGS. This does not contradict the previous statement, because the IQ estimate obtained from Piffer (2015) [9] was based on native Sri Lankans, since estimates for Sri Lankans living in the UK were not available. Given the moderate impact of environment, it is likely that the IQ of Sri Lankans living in the UK is actually higher than that of native Sri Lankans.

  158. utu says:
    @James Thompson

    Lee also uses the European prediction to estimate the intelligence of an African sample. He can predict 11% of European intelligence, and 1.6% of African intelligence.

    Let suppose that there indeed exists a polygenic score PGS that can predict IQ of world population of individuals to within the twin studies based heritability, i.e., 50-70%. Then this PGS is a sum of two polygenic scores:

    PGS=PGS_piffer+PGS_x

    where PGS_x is constructed using SNPs that do not appear in Piffer’s PGS_piffer and possibly some nonlinear terms of SNPs that appear in Piffer’s PGS. PGS_x may consist of 10’s if not 100’s of 1000’s of SNPs to account for 50-70% heritability.

    Piffer claims that subpopulation averages of his PGS_piffer predict subpopulation averages of IQ or in other words that is highly correlated (r≈0.88) with , where denotes subpopulation average. His claim implies that averages of on subpopulations must be close to zero or constant between subpopulations, meaning that is the same for Japanese and Snegalese and Italians… The constancy of is what Piffer implicitly assumes. The validity and probability of this assumption is not discussed. But this is a very strong claim. Is it possible that the the averages of PGS_x that consist of 100’s of 1000’s of SNPs will add up to the same number for Japanese, Senegalese and Italians? Is it probable?

    Say that PGS_piffer according to Lee could explain 11% of variance among Europeans and only 1.6% variance among Africans. This would mean that PGS_x explains 59%=70%-11% variance among Europeans and 68.4%=70%-1.6% among Africans and yet the averages for Europeans and Africans would be the same. Does it sound plausible?

    Piffer has made no progress since his last flawed paper that we discussed here some time ago. He still is pushing the same thing. He can be congratulated on persistence and chutzpah to have audacity to stay on the message that is flawed. Who is his target audience? Certainly not the scientists who could see through his work. It is just a propaganda work targeting ignoramuses out there who merely operate on confirmation bias principle. For them Piffer is a new Messiah straight from Israel. He does not seem to be stupid. He clearly knows the trade yet he opted the course that irrevocably ruined his scientific career. Perhaps he never cared for his career? Perhaps his day job is in a different field? Where does this carelessness come from? A character flaw? The same flaw that lead him to promote the evident fraud Uri Geller?

    By stating “Monte Carlo simulations strongly suggest that the reported findings are not a fluke.” you seem to bought to his argument. But look at his Fig. 4. There are about 20 different polygenic scores that produce correlation larger than 0.65 and possibly few that produce larger than 0.88. And he only did 943 random selections out of only 2411 SNPs. There are millions of polygenic scores out there that can correlate with his data set at r≈1. I even could postulate that arbitrary sequence (n=26) of random numbers can be matches with a polygenic score that correlates with it at r close to 1 from among 10 millions of SNPs. His Monte Carlo and p-values estimate is BS, it is blowing smoke up… It means nothing.

  159. @res

    Piffer supposedly used IQ scores because they are less sensitive to cultural context and infrastructure. I bet Indian immigrants to the US have a strong culture of formal education which boosts their education relative to their IQ. And I bet their raw intelligence level isn’t as high as suggested by their education. I found this study on the IQ of Gujarat which suggests a low IQ for environmentally deprived ones: https://www.ncbi.nlm.nih.gov/pubmed/25231330
    It would be good if someone could find other estimates of Gujarati IQ.

  160. res says:
    @South Asian Dude

    No, they weren’t doing a prediction of individuals within a group. They took a sample of a million individuals along with their educational attainment to figure out what variants are responsible for educational attainment. So by nature, the cohort of a million is a training set. You can only make a prediction AFTER you have figured out the weights.

    I skipped to your second step. The GWAS is a necessary precondition.

    From the prediction side, my point stands.

    The way p-value works is the larger the sample, the lower the number. So the cutoff they use tends to be a very low number (to the power of -8 from what I recall). You look at something called a mahattan plot to look at the outliers.

    I know. Probably better than you do (and it is Manhattan plot). If you read the paper you might have noticed that Piffer tried different significance thresholds and found: “The maximal predictive power was reached by selecting SNPs which met the conventional GWAS significance threshold (P < 5 ×10−8), whilst picking higher significance SNPs reduced the predictive power (due to reduced number of SNPs)."

    This makes ZERO sense if said group has not been studied before. The effect size is not going to be the same as Europeans. Do you understand that?

    Yes. I have noted just that many times in this blog.

    Which returns us to the question at hand. Why does Piffer’s approach work so well?
    And see comments regarding “ZERO” below.

    Which again makes zero sense if said group has not been studied before.

    The statistical averaging will help as long as there is some signal present.

    And are you honestly suggesting that there is ZERO commonality between races in the genetic structure of IQ?

    No arrogance here. The bottomline is, garbage in, garbage out. The OP agrees that other groups need to be studied due to unique variants. You don’t seem to grasp this concept. Not my problem.

    Your followup at least indicates you have more of a clue than I guessed from your individual/group inanity in your earlier response to me. But that you can’t tell I understand your point and am asking questions a level beyond it says a lot about arrogance.

    Let’s return to what was actually said in the OP:

    Piffer accepts that an important limitation of polygenic risk scores based on European DNA is that they may miss other variants.

    We are to a large degree having a disagreement in interpretation (how do the other group issues affect the predictions? is the prediction glass half empty or half full?). Which returns us to my question.

    Given the limitations of his method, why does Piffer get such good results?

    • Replies: @South Asian Dude
  161. @Merculinus

    Lee et al. didn’t use 1.1 million individuals as a training set. They used a sub-group as training set and tested the results on other samples. You are totally ignorant both of theory and methods behind Lee’s work, and Piffer’s, and even basic statistics. Your nickname says blatantly that your attachment to your ethnic group is clouding your reasoning.

    The prediction analysis was done on only a few thousand samples. To quote:

    The constructed polygenic scores for European-ancestry individuals in two prediction cohorts: the National Longitudinal Study of Adolescent to Adult Health (Add Health, N = 4,775), a representative sample of American adolescents; and the Health and Retirement Study (HRS, N = 8,609), a representative sample of Americans over age 50. We measure prediction accuracy by the “incremental R2”: the gain in coefficient of determination (R2) when the score is added as a covariate to a regression of the phenotype on a set of baseline controls (sex, birth year, their interaction, and 10 principal components of the genetic relatedness matrix).

    Also see their supplementary tables. The N is given as more than a million. That is the number of samples used. 1.1 million is indeed the number.
    Then they used like a 1000 African Americans to make another prediction. In fact, they say that the prediction is not accurate. But somehow you seem to know more than Lee et. al.

    Because the discovery sample used to construct the score consisted of individuals of European ancestry, we would not expect the predictive power of our score to be as high in other ancestry groups7,26,27. Indeed, when our score is used to predict EduYears in a sample of African-Americans from the HRS (N = 1,519), the score only has an incremental R2 of 1.6%.

    So Lee et al admit that the predictive power goes down in African Americans.

    I don’t see why Africans should have more beneficial mutations, which is the shaky ground on which your attack on Piffer stands.

    This is simple logic. I count 675 alleleles with effect size >0 and 595 alleles with effect size <0 in Europeans. So by logic, Africans will also have more positive effect size then negative ones.

    You also have to explain how come Ashkenazi Jews have such a high PGS and high IQ, or Finnish people.

    I don’t know how predictive PGS is for Ashkenazi Jews. Maybe it is, maybe it isn’t. The cohort was mainly Europeans. Maybe they are predictive for Ashkenazis. As for Fins, it is probably accurate. I don’t know to be honest. The study was done on Europeans, so Fins may really have higher genes for educational attainment.

    • Replies: @Merculinus
    , @j2
  162. @res

    I know. Probably better than you do (and it is Manhattan plot). If you read the paper you might have noticed that Piffer tried different significance thresholds and found: “The maximal predictive power was reached by selecting SNPs which met the conventional GWAS significance threshold (P < 5 ×10−8), whilst picking higher significance SNPs reduced the predictive power (due to reduced number of SNPs)."

    Okay fine, you know Manhattan plots and statistics better than I do. But hey, you are the one telling me that GWAS is based on individuals. Yet, p value is not an individual statistic.

    • Replies: @res
  163. @utu

    1) Piffer reached the same results in 2015 with only 9 SNPs. The fact that 4 years later a PGS built from 1.1 individuals, and comprising 2400 SNPs is almost identicaly to what he produced in his 2015 paper shows that the method goes a long way. This is called replication, something pretty rare in the social sciences. And this proves the constancy that you claim he’s assuming. If there had been no constancy, later GWAS with larger samples would not have confirmed his earlier results. His 2013 and 2015 predictions were that as more SNPs would be found, they’d conform to the same pattern.
    2) You obviously didn’t read the paper carefully or you can’t do math. The Monte Carlo simulations were carried out with 943 PGS, each comprising 2411 SNPs. How many SNPs are these in total? 2,273,573. Over 2 million! He didn’t do random selections from 2400 SNPs. He randomly selected 943 non-overlapping sets from over 2 MILLION SNPs!!! And no, no PGS had correlation with IQ larger than 0.88 (again this is written in the paper).The p value in this case is=1/943= 0.001.
    If you had the slightest idea of statistical significance you’d not make your idiotic claims. Piffer also provided the data and if you were serious you’d try to replicate results instead of spreading horseshit on this forum.

  164. @South Asian Dude

    You don’t know statistics and your logic is flawed. I said that if we remove the non African specific SNPs, we create a level playing ground because we remove the effect due to higher proportion of beneficial mutations among European due to population-specific variants. You would have to assume that the ratio of positive to negative effect among Africans is higher than that among Europeans, that is higher than 675/595. There is no reason to believe that this is the case.

    • Replies: @South Asian Dude
  165. @utu

    Thanks. There are some words missing in the third paragraph that you posted up, or something messed it up, and that has made it difficult for me to follow your argument.

    In paragraph four, I think that Lee argues that the European PGS is attenuated in predicting the African test group because of differences in linkage disequilibrium, not in actual SNPs. That is, the genes may be very similar, with lots of overlap, but finding the correct locations in African DNA on the basis of European DNA is difficult.

    • Replies: @utu
    , @utu
  166. utu says:
    @James Thompson

    words missing in the third paragraph – Indeed. I used brackets to denote averages and apparently the editor doesn’t like brackets. Here I rewrote it:

    Piffer claims that subpopulation averages of his PGS_piffer predict subpopulation averages of IQ or in other words that Avg{PGS_piffer} is highly correlated (r≈0.88) with Avg{IQ} on 26 subpopulations. His claim implies that averages Avg{PGS_x} on subpopulations must be close to zero or constant between subpopulations, meaning that Avg{PGS_x} is the same for Japanese and Snegalese and Italians… The constancy of Avg{PGS_x} is what Piffer implicitly assumes. The validity and probability of this assumption is not discussed. But this is a very strong claim. Is it possible that the the averages of PGS_x that consist of 100’s of 1000’s of SNPs will add up to the same number for Japanese, Senegalese and Italians? Is it probable?

    • Replies: @Merculinus
  167. @Merculinus

    You don’t know statistics and your logic is flawed. I said that if we remove the non African specific SNPs, we create a level playing ground because we remove the effect due to higher proportion of beneficial mutations among European due to population-specific variants.

    And I am saying that it is still not a level playing ground if we don’t know the effect size in African populations. To give you an example, light skin in Europeans and light skin in East Asians are due to different variants. So it is highly possible that the most predictive variants in Africans have not been discovered.
    Okay fine, I don’t know statistics. You do.
    We are going in circles here. No point in arguing anymore.

    • Replies: @Merculinus
    , @Merculinus
  168. @utu

    As I said, the constancy has been achieved, as this is a follow up from Piffer’s 2015 paper. The 2015 PGS used only 9 SNPs and predicted that constancy, that is that as more SNPs would be identified by GWAS, they’d conform to the same pattern. Now we have scores with 2411 and 3527 SNPs which are almost identical. This study’s PGS is already a PGS_x and Piffer’s 2015 is the actual PGS_piffer ( to use your notation) and includes N= 2,411-9= 2402 and 3,527-9=3518 SNPs.
    So you’re waiting for what has already been achieved. I can’t help it if you’ve not kept up with the literature from the past 6 years nor read Piffer’s earlier papers.

    • Replies: @utu
  169. utu says:
    @James Thompson

    I have an idea how Piffer could add an extra bit of independent verification to his methodology. But just one extra bit. He could use Lee’s 1,1 million data base to create quasi-Japanese, quasi-Italian and so on populations. But not by grouping individuals to produce the desired average PGS which would be easy but by grouping individuals that produce the same frequencies of SNPs as Japanese, Italian and so on populations. Piffer calculates his average PGS using frequencies f1,f2, f3,… of SNPs used in the polygenic score. He would have to come up with an algorithm that from among 1,1 million individuals finds the largest subset that has the same frequencies f1,f2, f3… as, say, Japanese have. Is there a guarantee that the subset is not empty? No but there is chance it may exist. The subsets could actually overlap since they are synthetic quasi populations. Then he would calculate the average educational attainment for each subset and correlate them with average polygenic scores that would be identical to ones he used for real populations. So he could make a plot just like Fig. 2 where values on x-axis would be the same but values on y-axis would be average educational attainments from Lee’s data set. What would be the correlation? Close to his r=0.88 or much smaller? And also he could correlate the educational attainment of quasi populations with IQ of real populations. If r was much lower than 0.88 he would have to admit that his approach is flawed. If however r was close to 0.88 he could argue that his approach works on quasi-populations where you make the explicit transition from GWAS predictor function for individuals to that of averages. But obviously this still would be insufficient to validate his approach. Just one extra bit of hand waving or perhaps a wake up call.

  170. utu says:
    @Merculinus

    You must not be understanding something. PGS=PGS_x+PGS_piffer suppose to predict 50%-70% of IQ variance in global population. Nobody has found such a predictor yet, so PGS_x is unknown, so nobody can prove that its average for different populations is the same. But Piffer implicitly postulates that it is so. One can play tricks (*) with postulates. Apparently the trick worked on you.

    (*) Einstein postulated that c is invariant in order to derive Lorentz transforms which he was implying that he did not know. But his bet was safe because Lorentz transforms imply that c must be invariant and Lorentz transforms were known earlier and studied by Lorentz and Poincare. They were derived from invariance of Maxwell equations by Lorentz about five years earlier. Einstein’s postulate could have been just a trick to hide the fact that he knew of Lorentz transforms. It worked for Einstein.

  171. @res

    You do not seem to understand my mathematical speak and the dynamics of the population in question.

    Intelligence trait in a free population should usually be roughly normally distributed.

    Notation presented a hypothetical population having the properties described below:

    a population had a more bimodal distribution where the right side of the curve was “normal” for an average IQ of 100 but the left side was distorted lower through missing a critical nutrient then the actual average might be lower even though the distribution of high IQ outliers was the same as if the average was 100.

    Let us simulate one (a hypothetical one though) out of a set of dynamics that would produce such a population distribution.

    Imagine that a country of same racial ancestry had only two states, A and B. Both states were identical in both population size and population mix such that each state contained half the country’s total size of smart fractions. Then a dictator emerged and erected a barrier between the states so that the populations do not mix.

    Next, the dictator subjected state A to extremely severe hardship so that its population experienced severe prenatal malnutrition. The dictator family lived for many generations.

    In a future generation, an intelligence test was administered to states A and B as a country, and the results were collated together.

    We would see a bimodal distribution, with overwhelmingly most of the smart fraction coming from state B. State A would have lost most of its “smart fraction” (by standards of the combined population) to impairment. The bimodality would occur if and only if both states have a common standard deviation sd, and the difference between the means of the distributions is greater than 2sd.

    Now image if state A was never discriminated against in the first place. They would have contributed more “smart fraction” candidates increasing the total size of the country’s smart fraction. Let us consider the country’s total smart fraction size under this ideal condition as s.

    s1 shrinks because of impairment.

    This refers to that loss in state A under the unideal condition.

    Such that (s1 + s2) < s.

    That refers to the composite deficit in total smart fraction size in the unideal country, with s2 being “smart fraction” contribution of state B.

    In the above state, we can say that the size of the smart fraction has been reduced.

    This holds true in the unideal country. There is a reduction.

    P.S. Conditions like this may happen in a real population due to, say, severe systematic withdrawal of “opportunity” from sections of the population.

    • Replies: @res
    , @notanon
  172. res says:
    @South Asian Dude

    But hey, you are the one telling me that GWAS is based on individuals. Yet, p value is not an individual statistic.

    I’m not sure how you got that from what I have written here. I’m not even sure what you mean.

  173. res says:
    @Chimela Caesar

    At least one of us is not understanding things here. I think notanon’s original comment and my elaboration were simple ideas and quite clear. The best interpretation I can come up with for your comment is that you are making something simple far too complex.

  174. j2 says:
    @South Asian Dude

    “” You also have to explain how come Ashkenazi Jews have such a high PGS and high IQ, or Finnish people.””

    I don’t know how predictive PGS is for Ashkenazi Jews. Maybe it is, maybe it isn’t. The cohort was mainly Europeans. Maybe they are predictive for Ashkenazis. As for Fins, it is probably accurate. I don’t know to be honest. The study was done on Europeans, so Fins may really have higher genes for educational attainment.”

    Individual data sample do not show strong correlation with any PGS and IQ. Only the country averages fit nicely on a line. If it would be so that individual samples would show a clear PGS vs IQ correlation, then this kind of research would have a basis, but as the individual correlation is very weak, they can tune PGS to raise or lower any small group while still keeping high the correlation between one version of PGS with another version of PGS.

    I think what best shows the flaw in Piffer’s method is exactly Finns. In earlier PGS versions what Piffer used, the British and Finns got very much the same PGS. In the last the Finns went clearly ahead of the British. It is the same dataset, so PGS was manipulated to give this result. Obviously, the PGS was manipulated to give this result in order to get the Ashkenazi Jews to the place the researchers wanted to put them. That is, they selected PGS in order to give country/ethnic group averages that agree with the present consensus of IQ values. Then they show that to the whole data set the correlation of the new PGS with the older PGS is still very good, but this new PGS moves some groups to different places, showing that there is intentional manipulation of PGS.

    In the 1970s the Finnish IQ was estimated as 97, while Edward Dutton claims it is now 102/103. In this time there has not been any genetic selection to explain such changes, therefore the effect can only be environmental. IQ reflects also environmental things. It must not possibly be used as a way to tune PGS to give correct country/group averages. Finns do have somewhat more genes that are more common in Eastern Asians (from 4% to 15% depending on a measurement), thus if you tune PGS in such a way that East Asians get the IQ they should get, it naturally raised the Finnish PGS.

    As for Finnish IQ, the PISA results are very good, but there are reasons for it. One is the language. Finnish is a very good language in the sense that students know how to spell it (it is written very much as it is spoken and the choice of letters is small) and in the sense that the meaning of words is usually clear, unlike in languages like English, where the vocabulary includes words that nobody knows. In Finnish the new words are concatenation of known words, so their meaning is easy to understand. This language should give some advantage in Reading in PISA. As for Mathematics, it is so that what PISA measures is closer to that what the Finnish school teaches, application of math, not theory of math. You do not see the Finnish school math team doing well in math olympics because there are no special schools for the talented and therefore good students cannot learn the things that are needed in order to do well. But in PISA, the way it is taught gives a certain advantage.

    Piffer’s method does seem to me to have an ideological goal, not a scientific one. Apparently the top of the scale does not matter much, as long as European whites are not leading it, but the real target probably is to get the developing world down. Or, so I see this.

    • Replies: @Merculinus
    , @notanon
  175. @Anon

    You simply know this from the fact that South Asians in the US are maybe the most educated group out there.

    They are also extremely unrepresentative of their home countries’s average cognitive level.

  176. @South Asian Dude

    It has been shown by several papers that the causal variants for height and IQ and other polygenic traits are the same in different populations. Skin color is not highly polygenic like height and IQ, so you’re comparing qualitatively different phenotypes.
    I told you many times already that focusing on the more significant SNPs (which are more likely to be causal) and the putatively causal SNPs as identified by Lee et al., increases the Black-White gap. For your attack on Piffer’s thesis not to crumble, the following conditions would have to hold: 1)Some SNPs might be polymorphic only among Africans; 2) These polymorphic SNPs are on average more beneficial than detrimental (yes, mr. Naive, most mutations aren’t beneficial); 3) The proportion of beneficial to detrimental population-specific alleles is higher among Africans than Europeans, because the Black-White gap stands strong even after removal of the non-African specific SNPs, which penalizes Europeans, thus creating a level playing ground.
    I think all of this happening together is very unlikely.

  177. @j2

    You get the conspiracy theorist prize for the readers of this blog! Kudos!

  178. @South Asian Dude

    Do you realize how pathetic you’re getting? You are shooting yourself on the foot, and pretty badly! You are admitting the possibility that the genetic architecture of intelligence is as different between Africans and Europeans as for skin color. So you’re more of a racialist than me. You’ve just implied that the White-Black gap in intelligence can be as large as that in skin color! Amazing!

    • Replies: @notanon
  179. notanon says:
    @Chimela Caesar

    there’s an unstated assumption in my hypothetical which is that class already correlates with IQ.

    for example say you have a country where everyone has an adequate amount of ingredient x and a normal IQ distribution the unstated assumption is class and IQ are already correlated and 98%+ of the smart fraction are already on the right side of the curve.

    then say ingredient x becomes restricted by cost so only the right side can afford it then the right side of the curve would stay the same and the left side would become stunted.

    • Replies: @Chimela Caesar
  180. notanon says:
    @Merculinus

    i may be misunderstanding but i don’t think he’s arguing from a blank slatist position.

  181. notanon says:
    @j2

    In the 1970s the Finnish IQ was estimated as 97, while Edward Dutton claims it is now 102/103.

    https://www.ncbi.nlm.nih.gov/pubmed/3486110

    Endemic goitre of moderate severity was mainly found in the East of Finland still in the 1950s but the whole country was moderately iodine deficient. The daily iodine intake determined both from food consumption and from the urinary excretion in population samples was 50-70 micrograms being lower in the East.

    Today the iodine intake in Finland is about 300 micrograms per day, the highest in Europe.

    • Replies: @j2
  182. notanon says:

    i wouldn’t know how to do this but hypothetically…

    if
    – there’s an ingredient x which needs to be sufficient for someone to meet their genetic potential
    – an entire population with sufficient ingredient x creates a normal distribution
    then

    wouldn’t it be possible to create different curves based on various assumptions of class or regional anomalies in the sufficiency of ingredient x which could then be compared to existing data?

    this might clear up some of the apparent anomalies in IQ data.

  183. j2 says:
    @notanon

    “Endemic goitre of moderate severity was mainly found in the East of Finland still in the 1950s but the whole country was moderately iodine deficient.”

    They started adding jodine to salt in Finland in the 1950s and have done it ever since. The IQ measurement giving 97 as the average was from adults and made in the year 1979 (or 1977, I cannot remember for sure). I was born in 1956 and I never heard of a single case of goiter in Finland even in older people (though there are about 400 yearly in 5.5 million in Finland today). Instead, goiter is still very common in Central Europe, in countries like Austria where the average IQ is 100. Goiter hardly explains the measurement. A much more likely explanation is improved food and especially education.

    • Replies: @notanon
  184. @notanon

    Notation, these things are nuanced by nature.

    Even with your new update on both the assumptions to be applied and the dynamics at work in the population, the size of the smart fraction is still reduced.

    Here is the explanation.

    We are always to consider a future generation for the assessment after the restrictions (cost of x) are applied (we allow time for restrictions to work on the population).

    Now, let the total size of the smart fraction in the population, assuming no restrictions were applied in the first place, be s.

    Let the total size of the smart fraction in the population, with restrictions applied, be s1.

    On the ability of the low IQ segment of the original population to produce “smart fraction” (by standards of the combined population) candidates in the future generation, see excerpt below:

    If a person with a high IQ marries someone with a lower IQ, their kids could have most any IQ. The same is true for two high IQ parents (although their kids will tend to have higher IQs). And for two low IQ parents.

    This possible range of IQs is so wide and unpredictable because so many different genes are involved in IQ. And because the environment plays an important role too.

    https://genetics.thetech.org/ask-a-geneticist/intelligence-and-genetics

    So the low IQ segment of the original population does produce “smart fraction” candidates in the future generation (albeit few). With restrictions applied on this segment, fewer “smart fraction” candidates are produced in future generations due to impairment.

    Hence,

    s1 < s

    The size of the smart fraction would reduce.

    • Replies: @notanon
  185. notanon says:
    @j2

    They started adding jodine to salt in Finland in the 1950s and have done it ever since. The IQ measurement giving 97 as the average was from adults and made in the year 1979 (or 1977, I cannot remember for sure).

    iirc iodine effects cognitive development in children so i guess you’d need to know the age of the participants in the 1979 study as anyone older than (?) when it started wouldn’t have been effected whereas 40 years later most people would be.

    • Replies: @j2
  186. notanon says:
    @Chimela Caesar

    So the low IQ segment of the original population does produce “smart fraction” candidates in the future generation (albeit few).

    fair enough

  187. j2 says:
    @notanon

    OK, there was another IQ test, giving 99 in 1997, and then there is a new one (I think 2014, a standardization cohort for a Finnish IQ test) giving 101 and Edward Dutton gives it as 102/103. Even dropping the first test there is growth and this growth is not in any way related to genes, and if you put the 99 into Piffer’s diagram, then the present diagram matches poorly as the British have 100 by definition and Finns would have 99 by measurement, yet the PGS would predict that Finns have a higher IQ.

    Just notice this, regardless of if Piffer or anybody else has made any manipulation or not, the method of changing PGS and that the results of British and Finns clearly change by changing the PGS mean that the method allows manipulation of PGS. It is totally insufficient to give the correlation of different versions of PGS to the whole data. One should study how the country values change and what causes it.

    For instance, for Ashkenazi Jews there should be values for each PGS that Piffer has used, not only for the last. Only in this way one could in a convincing way show that the last PGS has not been selected to give the wanted result. If a research method allows manipulation, this issue must be addressed regardless of if any manipulation was done or not. We cannot trust the author. The result must be such that we do not need to trust the author.

    And about Ashkenazi Jews, I still found no answer to this that the paper from 2019 states that the PGS used was developed from the Wisconsin sample in 2018. It cannot possibly be the PGS Piffer used in other tests done earlier. Yet Piffer’s table lists the papers from where he got the results. Most papers are older than 2018, they cannot have the same PGS made in 2018 for Wisconsin. Using different PGS for different tests is incorrect, the results cannot be compared.

    • Replies: @notanon
  188. res says:

    I still found no answer to this that the paper from 2019 states that the PGS used was developed from the Wisconsin sample in 2018. It cannot possibly be the PGS Piffer used in other tests done earlier.

    Piffer uses various GWAS as the sources for his PGS. Then evaluates those PGS on various populations (e.g. Wisconsin Jews, 1000 Genomes) for which genetic data is available.

    So the different PGS from various GWAS and selection criteria can be mixed and matched with different studies giving genetic population data as desired.

    Here is some discussion from his 2017 RPubs about some different PGS he checked:
    https://rpubs.com/Daxide/279148

    Correlation between the intelligence PS and IQ,PS_Piffer2017,PS_Piffer_2017_162SNPs are r=0.496, 0.646, 0.497. Correlations between the intelligence-EA PS and IQ, PS_Piffer2017_9, PS_Piffer_2016_162SNPs are r= 0.877, 0.924, 0.835 These are lower than the correlations previously observed (Piffer, 2017)

    One of the major points here is that the current PGS (EDU3) gives results very similar to the earlier versions (e.g. the 9 SNP version).

    Here is some discussion of the PGS involved from the current paper:

    Piffer (2017) [15](also see supplementary material) identified 9 genomic loci that were replicated across three of the largest GWAS of educational attainment (EA) [16–18]. The same 9 SNPs were successfully used to predict genetic differences in cognitive ability between ancient and modern samples [19].
    In addition, the full set of 2411 genome-wide significant SNPs from the latest GWAS of educational attainment (henceforth, “EDU3”) [7] will be employed.
    Lee et al. (2018) [7] also reported a set of 127 putatively causal SNPs (posterior inclusion probability >0.9). These will be used to provide a polygenic score which is less subject to linkage disequilibrium-decay from a theoretical point of view (see discussion).

    Back to you.

    Most papers are older than 2018, they cannot have the same PGS made in 2018 for Wisconsin. Using different PGS for different tests is incorrect, the results cannot be compared.

    Hopefully the above makes clear why this is not so.

  189. res says:
    @j2

    The Dunkel et al study used a sample of 53 Jews, but the page
    https://emilkirkegaard.dk/en/?p=7680
    claims that there is an independent verification from the Wisconsin study with 153 Jews and links to
    http://rpubs.com/Jonatan/jewish_pgs
    Looking at the results of Jonathan Pallesen 2018 from the Wisconcinn study the number of Jews was also 53, not 153, and it is not any independent study. The figures of density vs IQ are exactly the same as in Dunkel et al, so this is not any independent verification. Pallesen is the third author in Dunkel et al and the data seems to be the same 53 Jews.

    Looking at your link of http://rpubs.com/Jonatan/jewish_pgs
    I see that the study with 53 Jews was the WLS (heading 2). The replication study Emil referred to in his page you linked was the HRS:

    While the paper was in review, we found another dataset, Health and Retirement Study (HRS), which also had the necessary variables. Unfortunately, the IQ test is worse. However, our main findings replicated (see the codebook). This dataset has 153 Jews with the right variables.

    The HRS is heading 3 in Jonatan Pallesen’s page. So Emil was actually referring to the HRS you mentioned. Not the 53 Jewish subject WLS as you asserted in “data seems to be the same 53 Jews”.

    Emil does seem to have made a mistake saying the HRS had 153 Jews. Per Jonatan Pallesen’s page the correct number is 212.

    Regarding

    Concerning the polygenic scores, the study in Pallesen divides PGS into two scores, which allows a selection of markers to get a result you want.

    Only for the HRS (PGS_EDU and PGS_COG). The WLS only gives a single PGS “polygenic scores for educational attainment (edu PGS)” which I think is the same as PGS_EDU in HRS.

    So only one PGS is common to both data sets and no arbitrary choice is possible. If you look closely at section 3 for the HRS you will see that PGS_COG is only used in the initial set of three plots. PGS_EDU is used for the rest of the analysis.

    If you are going to make vague accusations like:

    My conclusion is that Dunkel et al is even more suspicious than Piffer ever.

    Please try to do a better job on your analysis.

    P.S. In “In Pallesen’s study the Plot of PGS vs IQ entitled Religion and with the axis IQ and PGS do not have a normal distribution of IQ for Christians, therefore Christians comprise several groups.”

    I assume you are talking about the scatterplot in section 2 (WLS)? How do you see that the IQ distribution is non-normal? I find it hard to extract the IQ distribution from the scatterplot and since both it and the earlier distribution plot are using the same IQ variable they should be the same. Except the distribution plot is almost certainly smoothed, which may be the difference you are noting.

    • Replies: @j2
  190. j2 says:
    @res

    You sure sound like one of the authors, then if so, please answer to this:

    In the HRS study with a larger sample Jews and Christians differ only little in the cognition plots,
    but much more on the two PGS plots. This indicates that PGS includes something that gives higher scores to Jews while not being relevant for IQ. A similar observation can be made from the Plot of PGS vs IQ. In that plot the Jewish data points are spread to the right, but not especially up. It looks like the PGS picks up something that Jews have (frequencies of SNPs correlating with Jewish origins) but it does not correlate with IQ. The relation of the PGS to IQ seems to be similar for Jews and Christians.

    You have not answered to the question why the UK and Finn scores differ in a way that suggests manipulation and even if there is no manipulation, a paper should clarify the issue so that it is clear that there is no manipulation. National IQ scores differ in different measurements and there is no doubt that environmental issues affect them, these issues including nutrition and education, while Piffer’s PGS scores reflect gene frequencies and cannot be affected by environmental issues. Therefore finding a too good correlation of national IQ and PGS does in a natural way suggest that this PGS must be matched to give the result.

    In case you are an author of any of these papers, please note that an answer of this type:
    “If you are going to make vague accusations like:
    My conclusion is that Dunkel et al is even more suspicious than Piffer ever.
    Please try to do a better job on your analysis.”
    is totally inappropriate. It is the author’s task to make sure that all such issues are clarified. Both Dunkel’s and Piffer’s papers leave much to be desired. Their level of analysis is too poor.

    • Replies: @Merculinus
    , @res
  191. @j2

    Your thinking is manipulated, not the data. Simply because in a 2015 paper using 9 SNPs the rankings are slightly different than in the 2019 paper which used 2411 SNPs? Do you have the slightest ideas of what uncertainty and confidence intervals mean? Are you aware that a score built only from 9 SNPs is subject to a lot of random fluctuations?You are also cherry picking a discrepancy in a typically schizophrenic way. You omit that the correlation between 2015’s factor and the 2019’s is 0.96. A couple scores are different, so what? Go back to statistics 101.
    And this is me being generous. I just checked Piffer’s 2015 paper, and the Metagene score ranks Finns higher than British (table 2). The metagene score was computed via factor analysis, the “polygenic score” simply as an average, which is a viable approach when you have such a small number of SNPs. So you cherry picked two populations, but also one of the two polygenic scores, omitting the other one.
    Since when a result that is “very good” becomes “too good”? Should researchers stop publishing positive results, lest some schizo-conspiracy theorist suggest that they’re too good to be true? Would you rather see bad results? It is clear from your approach that you’ll always find something to complain about. Instead of making silly accusations, compute the PGS yourself and come back with the data. Until then, I will stop reading your comments.

    • Replies: @j2
  192. notanon says:
    @j2

    there is growth and this growth is not in any way related to genes

    well yes and no.

    i think the most likely answer to all this will be populations have a maximum genetic potential which can be stunted by the absence of one or more critical nutritional ingredients during childhood.

    so i agree in the Finnish case the improvement is probably not to do with an improvement in their genetics (cos too fast) just a removal of a dietary deficiency that was stunting them below their pre-existing genetic potential.

  193. j2 says:
    @Merculinus

    From your comment it looks like you are lacking 30 years of research after Ph.D. Come back after you have it.

    • Replies: @Merculinus
  194. Anonymous [AKA "jp7"] says:

    In Pallesen’s study the Plot of PGS vs IQ entitled Religion and with the axis IQ and PGS do not have a normal distribution of IQ for Christians

    Yes it does.

    Looking at the results of Jonathan Pallesen 2018 from the Wisconcinn study the number of Jews was also 53, not 153, and it is not any independent study. The figures of density vs IQ are exactly the same as in Dunkel et al, so this is not any independent verification. Pallesen is the third author in Dunkel et al and the data seems to be the same 53 Jews.

    Yes. The 53 Jews from WLS are both in the paper and in the RPub. Nothing wrong with this.

    Then there is a replication study mentioned in Pallesen with 212 Jews (3. Replication in HRS). The cognition scores differ very little and it points to an IQ difference of 3 points.

    As noted in the RPub, the IQ test in HRS is of very low quality. So it is not clear how large an IQ difference you would expect to find between the groups with this measure.

    Concerning the polygenic scores, the study in Pallesen divides PGS into two scores, which allows a selection of markers to get a result you want.

    @res covered this one.

  195. @j2

    The difference is I debunked your disgusting conspiracy theory, yet you can only reply with ad-hominem attack. Come back after you can provide the data to back your conspiracy theory, instead of this nonsense. Data manipulation is a serious allegation, and serious allegations require serious research, something which you obviously lack the tools to do.
    I wonder what kind of PhD plus 30 years of research you have. I had no idea they awarded PhD in conspiracy theory.

    • Replies: @j2
    , @j2
  196. res says:
    @j2

    You sure sound like one of the authors, then if so, please answer to this:

    I’ll take that as a compliment, but it is very incorrect. I am not even in the field. Just a dilettante engineer with an interest in individual differences, genetics, and IQ in particular.

    That statement does help me calibrate your judgment though.

    In the HRS study with a larger sample Jews and Christians differ only little in the cognition plots, but much more on the two PGS plots.

    That does seem odd. I attributed it to the inadequacy of the IQ measure used in the HRS (which was mentioned at both links).

    You have not answered to the question why the UK and Finn scores differ in a way that suggests manipulation

    Outliers happen. Especially when there are so many variables in play (see my other comment about why I am surprised the results are so good). I haven’t looked at that particular issue in any detail.

    Therefore finding a too good correlation of national IQ and PGS does in a natural way suggest that this PGS must be matched to give the result.

    I agree that is (and should be, IMHO) one of the first thoughts that occur when looking at these results. See the other Piffer threads for comments along those lines. Here are three (from Dr. Thompson, utu, and myself) which raise the issue of “too good to be true.”
    http://www.unz.com/jthompson/a-piffer-pause/#comment-1902335
    http://www.unz.com/jthompson/genetics-of-racial-differences-in-intelligence-updated/#comment-1897662
    http://www.unz.com/jthompson/tilting-at-sex-differences/#comment-2231250
    The discussions following those (and other comments in the threads) might be helpful.

    AFAICT though, Piffer is not doing that. The most recent work uses the Lee paper PGS (both full and causal versions). In general Piffer seems to be
    1. Using the SNPs supplied by others in their papers.
    2. Being open about the results when he tries new sets of SNPs (see my comment above including both IQ and EA PGS results, even though the IQ PGS results were not nearly as good).

    In short, to my eye he is acting exactly like a person with a good result who is making an honest attempt to validate it through replication (and address issues raised like linkage disequilibrium and spatial autocorrelation) would.

    In case you are an author of any of these papers, please note that an answer of this type:
    “If you are going to make vague accusations like:
    My conclusion is that Dunkel et al is even more suspicious than Piffer ever.
    Please try to do a better job on your analysis.”
    is totally inappropriate.

    Perhaps, but as a third party I think my comment is both appropriate and exactly on target. If anything is inappropriate here it is you making accusations based on little or no evidence.

    It is the author’s task to make sure that all such issues are clarified. Both Dunkel’s and Piffer’s papers leave much to be desired. Their level of analysis is too poor.

    Two things come to mind here.

    1. “Don’t let the perfect be the enemy of the good.”

    2. Isolated demands for rigor: https://slatestarcodex.com/2014/08/14/beware-isolated-demands-for-rigor/

    Do you apply this standard to all the papers you read? If so, I sincerely hope you don’t read very many Narrative confirming psychology papers (e.g. stereotype threat).

    FWIW, if you want to make a serious effort to critique this work (and I think it could use some good critical thinkers with open minds going over it) please try to do the following.

    1. Improve your analyses.
    2. Be a bit slower to accuse people of malfeasance. You know too little about Piffer’s replication efforts to make a serious criticism of them. And I have a multi-year comment history here which should make clear I am not a researcher in the field.

    • Replies: @j2
  197. j2 says:
    @Merculinus

    “Merculinus, I wonder what kind of PhD plus 30 years of research you have. I had no idea they awarded PhD in conspiracy theory. ”

    Right, you are really brilliant, Merculinus, but I would not choose you as my Ph.D. student because your logic is all false, indicating that you are too stupid for a Ph.D.

    I would not discard you for arrogance. We in the academy always tolerate arrogance as it is a common male mistake and made often by quite, even very, intelligent people, but your shortcomings are in the basic logic, your logic is faulty, that’s it, I cannot help it and you are rude, which does not help. So, I would not take you as a Ph.D. student never, if you write as you write, nor would any professors I know, I mean the clever ones. But maybe you know professors who would (the corrupted ones), not so strict on minimum level of intelligence for the Ph.D. I know, there are such professors. They serve one ethnic group. No need to identify the group, but it is not Finnish, nor French, German, Icelan or any other such a group.

    Anyway, I liked your suggestions that I might be schizophrenic. Well, I am not, and no student claiming so would be accepted for doctorate studies, those claiming so are simply schizophrenic, exactly like you, Merculinus. When did you make your Ph.D. and on what? You see, I made my Ph.D. in a real scientific field 30 years ago and you are simply a beginner, and in a highly dubious field. As you, I would simply stay quiet.

    • Replies: @Merculinus
  198. j2 says:
    @Merculinus

    “I wonder what kind of PhD plus 30 years of research you have. I had no idea they awarded PhD in conspiracy theory.”

    You are exactly right, they do not award a Ph.D. from conspiracy theory, nor do they award professorships, so maybe we can agree that I have all the merits and you have exactly none, but you are a bigmouth. The questions I made to the authors of the papers are all valid, the emotional reaction from you was not valid at all. It was simply an emotional reaction of a person who has no scientific qualifications, so to be ignored.

    But I do not ignore anybody, even you. You can present your case and I will think about it carefully, This is much more than what you ever would get in science, but I agree to do it. Despite your arrogance, I will do it, a fair trial, your logic against mine. Fare enough? I am always fair and never call unknown people schizophrenic, but maybe you come from some very rude an offensive subculture. I give you all the chance though I already can see that intellectually you will fail the challenge. Your people always fail it.

  199. j2 says:
    @res

    OK, res. You are not an author.

    If I would put my paper in some forum, certainly I would defend it, but this paper is not defended by anybody. But I would defend my paper under my name, naturally, all researchers should do it.

    So, you are not one of the authors, so I have no comments to you. It is simply so that I, as a reader, concerning e.g. claim that there is manipulation in the PGS. It is not for me to prove that there is manipulation, IO have not claimed anything, it is for the author of the paper to prove that there is no manipulation, he has claimed something.

    I exactly know that you are one-two generations younger than me and there has been all this brainwashing, so how could you possibly know what scientific truth means. You do your way, as dolls pulled by some doll master byu strings, but I guess it has to be so today. Maybe when you, res, grow a bit older, you will understand things you do not understand today. (Please, do not reply before you have reached my age and have some minimum level of wisdom. You are incredibly stupid now, only you do not see it, nor can anybody tell it to you, but I can see it, I cannot help you now, just trust me on this)

    • Replies: @res
  200. res says:
    @j2

    I would defend my paper under my name, naturally, all researchers should do it.

    Piffer did that here in the beginning. The best example of that was this post and the resultant thread:
    http://www.unz.com/jthompson/piffer-replies-to-prof-posthuma/

    I think his letter to Dr. Posthuma (which made up the whole of the blog post) is a good example of defending one’s work.

    After that thread I can understand why he hasn’t continued commenting here and seems to be focusing on refining his work. Which is the more important task anyway.

    So, you are not one of the authors, so I have no comments to you.

    So my reasoned comments don’t matter, just who I am. Thanks for making clear that is how you think. I don’t even have a PhD (in a related OR unrelated field) so I must know nothing, right?

    I hope you were really good at your specialty, because you don’t seem very skilled at either understanding material outside of your field or reasoned argument.

    I exactly know that you are one-two generations younger than me and there has been all this brainwashing, so how could you possibly know what scientific truth means. You do your way, as dolls pulled by some doll master byu strings, but I guess it has to be so today. Maybe when you, res, grow a bit older, you will understand things you do not understand today. (Please, do not reply before you have reached my age and have some minimum level of wisdom. You are incredibly stupid now, only you do not see it, nor can anybody tell it to you, but I can see it, I cannot help you now, just trust me on this)

    That was an epic ad hominem. Perhaps some other time you can back it up with something of substance.

    And BTW, thanks for continuing to make clear how bad your judgment is.

    P.S. What is it with people who think calling me stupid is some kind of decisive statement? I think my comments speak for themselves in that regard.

    • Replies: @j2
  201. j2 says:
    @res

    “That was an epic ad hominem. Perhaps some other time you can back it up with something of substance.”

    Whatever, res. You know, I am a quite old, retired professor from three universities. I do not really know what you young guys are trying to do, maybe you got mixed up. It is simply so that if you publish a paper, it is you, as the author, who has to show that the claims you make are true. It is not so that a reader has to demonstrate that there is an error.

    When there were religious texts, it was indeed so that they were true unless somebody could demonstrate that they are false, but this is not so in science. Piffer and Dunkel et al have to explain what PGS is used and how it relates to each ethnic group. All my comments were normal and allowed comments in science concerning a published paper. You and merculinus, not being researchers and not knowing what is the scientific practice, came to the defense of the authors, totally in vain. What I asked is the correct thing to ask, what you declared were not correct things to declare.

    I have not made an ad homenim attack against you. Either you have academic qualifications, or you do not have them. You do not, I do. That’s it.

    • Agree: Okechukwu
    • LOL: res
    • Replies: @res
  202. res says:
    @j2

    I have not made an ad homenim attack against you. Either you have academic qualifications, or you do not have them. You do not, I do. That’s it.

    For all of your eminence you don’t seem to know what an ad hominem is. Time for some remedial work. Hopefully this old dog is still capable of learning new tricks, but I guess we will see.

    https://en.wikipedia.org/wiki/Ad_hominem
    http://paulgraham.com/disagree.html

    Ad hominem (Latin for “to the person”[1]), short for argumentum ad hominem, is a fallacious argumentative strategy whereby genuine discussion of the topic at hand is avoided by instead attacking the character, motive, or other attribute of the person making the argument, or persons associated with the argument, rather than attacking the substance of the argument itself.

    You attacked my lack of “academic qualifications” (you mean like a PhD in an unrelated field? I wonder what actual experts in this field think about your assertion of qualifications here) rather than attacking the substance of my arguments. That is very definitely an example of the ad hominem fallacy in action.

    I mean really. It is a good enough example I think I will be referring back to this comment of yours in the future as one of the most blatant examples of an ad hominem I have ever seen in the wild. Just reread my quote from you above. And not even an attempt to engage with my substantive points. That quote is also one of the best examples of a self refuting comment I have ever seen.

    In Graham’s hierarchy of disagreement this comment of mine is what is known as refutation. You might try it sometime.

    P.S. For bonus points look up the appeal to authority fallacy:
    https://en.wikipedia.org/wiki/Argument_from_authority

    P.P.S. I think you have misread Merculinus rather badly as well (not surprising given your demonstrated poor judgment throughout this thread). I would recommend reading comment 130. But I will leave it to him to respond for himself as he sees fit.

    • Replies: @j2
  203. j2 says:
    @res

    About the lack of scientific credentials, let us notice that your friend Merculinius claims I have schizophrenia, so I simply pointed out that his comment shows that he is not an academic researcher, as no professor or serious researcher would write like that.

    About you lacking academic credits, that is not important. What I meant is that if you are not the author, your apologies to the paper can be ignored. That is not an offense, do not take it like that. It is even much kinder wording than what your friend Merculinius uses, it is simply that you do not know and you write, one should not do that. Your apologies must be ignored as you mix up things. That is, you do not know and yet you write. You explain that in the explanation that Dunkel gave to the 2018 paper the independent study was the HRS study, but this is false as it was not 153 Jews but over 200 Jews in that study. For such reasons as this I do not consider your apologies for Piffer and Dunkel valid. Your apologies are irrelevant to the evaluation of the papers, as they are false and you are not one of the authors.

    “P.P.S. I think you have misread Merculinus rather badly as well (not surprising given your demonstrated poor judgment throughout this thread). I would recommend reading comment 130. But I will leave it to him to respond for himself as he sees fit.”

    “130. Some colleagues have gotten access to an African sample and validated the Lee et al. SNPs on them. Simple.”

    What special do you see in this comment (except for that it should be have got, not gotten), except for nowhere demonstrated claim by Merculinius? For sure it does not explain why Merculinus thinks he can do a schizophrenia analysis and why he uses such offending language as he does. And why do you write things you write? What is wrong with you two? Like this:

    “I mean really. It is a good enough example I think I will be referring back to this comment of yours in the future as one of the most blatant examples of an ad hominem I have ever seen in the wild. ”

    So, why are you upset? Just go to make a Ph.D. and that’s it. It is some work, but very much easier today than at the time I did it, especially if you do it from a soft field like IQ.

  204. @j2

    I won’t deal with your ad hominem attacks anymore, but instead focus on your silly allegations.
    Reading your replies to RES also shows how deranged your thinking is.
    You are the one that has to prove that Piffer manipulated the data. You obiously ignore the basic principle of presumption of innocence. I suggest you look it up. It holds for science as well. A colleague of mine once felt that his work had been plagiarized and complained with the journal’s editor. The editor asked him to submit a paper that would prove that plagiarism had taken place, and this paper would be peer reviewed. The reviewers found some evidence of plagiarism, but since this was not prove beyond reasonable doubt, his allegation wasn’t published.
    Here we are not asking you to submit to the strict standards of peer-review, we’re asking you to at least do some work and show how exactly Piffer manipulated his data. All you can come up with are unsubstantiated claims, and you’re too lazy or arrogant or incompetent (or all of these things) to bother providing any evidence for your claims.
    Piffer obtained the polygenic scores from the supplementary data provided by Lee et al., and his results were very similar using EA_MTAG, CP_MTAG, and EA SNPs, at different significance thresholds, and using the full set, he used weighted and unweighted scores…what else do you want?
    He did not cherry pick SNPs here and there. Unlike you, I’ve read his full paper many times. You probably didn’t go past the abstract or the introduction.
    I invited you to try and reproduce Piffer’s results. Go ahead and compute the Polygenic Scores using Lee’s various GWAS hits, and see if he actually omitted some SNPs to get his fabulous results. Or download his code and data and see if it there are any errors in the code or if the data he used altered Lee et al’s original data. I checked all of this myself and I didn’t find any evidence of data manipulation. In the old days, readers had to take results at face value. Nowadays, with open data, you can actually work with the data yourself and see if there are some inaccuracies. Obviously your laziness or lack of skills is preventing you from doing that.
    Piffer provided data and code, now it’s your job to prove that he altered the data.
    Do you want us to believe your allegations simply because you’re old and you have a PhD (assuming you aren’t making this up)?
    Relying on an academic qualification on an unrelated field plus old age alone aren’t sufficient to prove your point.
    Your beef against Piffer’s results are based only on these two points: 1)His results are too good. 2) Some of the population scores change when he adds more SNPs. As to 1, if I were Piffer I’d take this as a compliment. Good results are good, bad results are bad.
    2)If you had a basic grasp of statistics you would not be surprised by this phenomenon. Go back to statistics 101 (if you ever took that course) and study/do homework on measurement errors. Try to measure your height, or your IQ (assuming you have an IQ) three or four times, and when you get different scores, please don’t come back to this forum claiming that the IQ tests, or the tape were manipulated. Yes, polygenic scores change as you add more SNPs, and this is measurement error. It is not necessarily the product of some evil force or biased author trying to make his theory look good.
    Again, if you want us to believe your claims, go ahead and show which SNPs Piffer omitted, and how adding those omitted SNPs would worsen his results.

    • Replies: @utu
    , @j2
  205. Anonymous[214] • Disclaimer says:

    So now we know j2 is a moon-landing cospiracy theorist. I guess Merculinus’ diagnosis was spot on.
    Here is what he wrote (http://www.unz.com/comments/all/?commenterfilter=j2) “But the issue remains that some of the photos are forgeries. Making composites and claiming they are not composites is forgery. In both science and the court room if your evidence is found to include forgery, your demonstration fails simply because of that. Nobody looks further and tries to find some true bits of information from your demonstration. Therefore, NASA claim of landing a man on the Moon fails.”

    • Replies: @j2
  206. utu says:
    @Merculinus

    go ahead and show which SNPs Piffer omitted, and how adding those omitted SNPs would worsen his results

    He omitted 10’s or 100’s of 1000’s of SNPs which haven’t yet been found that could explain 50-70% of IQ variance. And they haven’t been found not because of the lack of effort.

    Hsu could explained only 9% of variance using close to 10,000 SNPs and by adding another 40,000 SNPs he could not improve it. Then Hsu wrote a paper on a non-linear polygenic score function that he hoped would close the heritability gap. Either he closed it and did not bother to tell us about it (perhaps CCP took the results to transform the Chinese Communist Party Central Committee into a super race) or he did not close it and fell silent. His gung-ho attitude got somehow deflated.

    And look at the fabulous work of Lee with his 1,1 million individuals data base. What did he actually accomplish? With GWAS he identified 1,271. SNPs:

    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).

    As we see the truth is, which does not jump at you form the main body of Lee’s paper (or from J. Thompson article about it), is that to obtain 11-13% prediction each “lead SNP” from the “treasure trove” of 1,271 had to be combined with additional 787 SNPs on average. ONE MILLION SPNs in the polygenic score is a lot. Their polygenic score uses 10% of all SNPs there are in human genome and yet it can explain only about 10% of the phenotype variance. Are we closing the heritability gap?

    So, what is the state of research? The SNPs identified by GWAS (the treasure trove), i.e., the “causal” SNPs can predict 3.9% of variance only. To push this prediction to around 10% you need add 1000’s more of SNPs as done by Hsu who used a brute force lasso method and did not waste time on intricate GWAS or by Lee who ended up using circa 1 million of SNPs in his polygenic score.

    The bottom line is that we have no genetic predictor of IQ or education attainment. The heritability gap remains wide open. From GWAS we know of couple thousands of SNPs that on their own can predict only 3.9% of variance, yet we want to predict 50-70% of variance if we want to close the heritability gap. And at this point enters Mr. Chutzpah who happens to find that using the polygenic score of 9 and then of couple thousands of SNPs correlates (r=0.88) with IQs of 26 countries. And Mr. Chutzpah claims that it must be causal not spurious. Perhaps he was encouraged by J. Thompson who tried to create a meme that correlation means causation not so long ago here at unz.com.

    https://www.unz.com/jthompson/correlation-is-not-causation-but-its-the-way-to-bet/

    Perhaps the Correlation is Causation is a motto of new school of epistemology founded by Richard Lynn but so far this hasn’t gained acceptance in the main stream science.

    The bottom line is there is no reason to think that Piffer result is not spurious. At this point there is no reason to even believe that the heritability gap will be closed. Piffer’s result bets against low probability that the remaining SNPs that are not included in his PGS produce averages that are constant across 26 countries he selected. The validity of Piffers result hinges on a hidden assumption that the polygenic score of remaining SNPs (which can be 100’s of 1000’s of them) produce the same averages for different populations (countries, ethnic groups, races). This assumption remains a postulate that has not been proven and there are no indications that it could or should be true.

    Mr. Chutzpah is good for true-believers and wishful thinkers like J. Thompson, res and yourself who subscribe to the confirmation bias epistemology of Richard Lynn. You have to face it that you with your Mr. Chutzpah are at the level of cargo cult science. You think that if you mimic some scientific behaviors like calculation of P-Value and adorn it with scientific jargon that you are actually doing a science. You are on the level Melanesians who are building a plane using bamboo sticks in a hope that this ritual will cause materialization of boxes with cans of Spam. You even create a journal ‘psych’ that is a part of this cargo cult to bring the cargo cult mimetic behavior to even higher level so more people can be fooled. No Spam for Davide Piffer.

    • Agree: Okechukwu, AaronB
    • LOL: CanSpeccy, res
    • Replies: @j2
    , @j2
    , @Merculinus
    , @j2
    , @AaronB
  207. j2 says:
    @utu

    You are quite correct, utu.

    Concerning scientific papers, authors are not assumed correct until shown incorrect. A paper being peer-reviewed in some journal does not imply that the paper is correct. As I do not claim that the PGS is calculated incorrectly, I do not need to recalculate it. I claim that the PGS is selected in a way that has a methodological error. I will show what the error is:

    Ashkenazi Jews have genes that reflect a different ancestry and therefore they can be distinguished form the earlier groups in Piffer’s study by selecting to the PGS some SNPs that reflect Ashkenazi ancestry.

    A new PGS is formed in order to include Ashkenazi Jews to Piffer’s study. This new PGS is calculated partially from self-estimated mathematical talent and partially from educational achievements in Wisconsin. If Ashkenazi Jews over-estimate their mathematical talents more than other people in Wisconsin, or if they have higher educational achievements than the other people in the Wisconsin study, then a PGS formed with these criteria will give higher scores to Ashkenazi Jews than to the average people in the study. That is, this is the goal of the PGS: to give higher scores to those who have higher scores in educational achievement with the criteria used.

    The error is here: because the correlation with these genes and educational achievement is very weak on a person level, the PGS selected in this way will very probably reflect Ashkenazi ancestry and not only educational achievement. The selection method cannot differentiate between these two. The correct way would be to exclude Ashkenazi Jews from the creation of the PGS, select the minimal PGS without any such criteria that might favor some group (like self-estimated talent, which favors a group that most over-estimates its talents), and then to use this PGS to Ashkenazi Jews. But this is not done so.

    A further methodological error in both Dunkel and Piffer is that they do not study what are scores for other groups which have similar ancestry as Ashkenazi Jews, i.e., for instance Mizraim Jews or some Mediterranean/Middle East populations. If the scores for these groups do not show the expected correlation with educational achievements, then the PGS is poor. However, this is not investigated in either paper. Therefore the new PGS is not validated.

    A further methodological error is that the group of Ashkenazi Jews is too small. It is not too small for calculating the score, but it is too small in the sense that the new PGS can give any wanted PGS score to Ashkenazi Jews without much changing the correlation of the new PGS with earlier PGS versions Piffer had used when the correlation is calculated over all used ethnic groups. If there would be a very large sample of Ashkenazi Jews we might see that the new PGS does not measure educational achievements as well as the older ones, but this cannot be seen because the Ashkenazi Jewish group is so small that the effect does not show up.

    These are methodological errors regardless of whether they have caused incorrect results or not, but there are further reasons to suspect that they have caused errors:
    1. In the plot Plot of PGS vs IQ in Dunkel’s study the data points for Ashkenazi Jews seem to be spread more in the horizontal direction than in the vertical direction. That would be expected if the PGS reflects Ashkenazi ancestry and not only educational achievements.
    2. The same can be seen in the HRS study: the difference in cognition is small but the difference in the two PGS plots is large. This is what would be expected if the PGS reflects Ashkenazi ancestry and not educational achievements.
    3. Piffer did not calculate PGS scores to Ashkenazi Jews in his earlier studies, that is, earlier versions of his PGS, like the first one with 7 SNPs. This makes one suspect that the reason was that Ashkenazi Jews scored close to Italians and not to the high end where Piffer wanted to have them. One could check this, I will not care to do it as I am not reviewing the paper. It is likely to be so. Otherwise it is hard to understand why a new PGS was needed for Ashkenazi Jews and why it was made from the Wisconsin sample where Ashkenazi Jews were included: that is, why it was made in a way that has a clear methodological error.

    This is what you write:
    “Your beef against Piffer’s results are based only on these two points: 1)His results are too good. 2) Some of the population scores change when he adds more SNPs. As to 1, if I were Piffer I’d take this as a compliment. Good results are good, bad results are bad.
    2)If you had a basic grasp of statistics you would not be surprised by this phenomenon. Go back to statistics 101 (if you ever took that course) and study/do homework on measurement errors. Try to measure your height, or your IQ (assuming you have an IQ) three or four times, and when you get different scores, please don’t come back to this forum claiming that the IQ tests, or the tape were manipulated. Yes, polygenic scores change as you add more SNPs, and this is measurement error. It is not necessarily the product of some evil force or biased author trying to make his theory look good.
    Again, if you want us to believe your claims, go ahead and show which SNPs Piffer omitted, and how adding those omitted SNPs would worsen his results.”

    The beef is not in those things. One issue is that there is a methodological error in the way Ashkenazi Jews are included. Another is that there is a methodological error in Piffer’s way of plotting national averages from data that should not give such nice results except for by coincidence or by selection. Utu wrote about this problem, i have nothing to add to that.

    You write like this:
    “I won’t deal with your ad hominem attacks anymore, but instead focus on your silly allegations.
    Reading your replies to RES also shows how deranged your thinking is.”

    But it is you who started writing this kind of text:
    “You get the conspiracy theorist prize for the readers of this blog! Kudos!”
    ” Do you have the slightest ideas of what uncertainty and confidence intervals mean? Are you aware that a score built only from 9 SNPs is subject to a lot of random fluctuations?You are also cherry picking a discrepancy in a typically schizophrenic way. Go back to statistics 101.”

    • Replies: @Merculinus
  208. j2 says:
    @utu

    Sorry, utu, I replied to your comment when I was to reply to Merculinius. Your comment is very correct. Ignore the reply to Merculinius.

    • Replies: @utu
  209. j2 says:
    @Anonymous

    “So now we know j2 is a moon-landing conspiracy theorist.”

    I am not a Moon Hoaxer. I am Moon Hoax agnostic. No clear evidence either way.

  210. @utu

    According to you, Piffer’s study should be discarded because it doesn’t include all the SNPs in the DNA. With your logic, pretty much all the papers published in biology and psychology would be discarded, because they are based on samples from a population, and do not take measurements from all individuals. Polls and surveys are useless too. Climate measurements are inadequate because they don’t take temperature from every possible spot on the planet and chemical analysis of water is wrong because it relies on tiny bottles, instead of bringing to the lab all the water that is contained in the oceans, lakes, and rivers. Same goes for blood analysis: I will tell my doctor that I don’t trust my blood results. Next time he’ll have to draw all my blood out, and send the report to my grave.
    Well, I am sorry but you won’t get the Nobel prize, because the ideas of measurement error and statistical significance were developed a long time ago, to deal with these issues.
    Kudos! You have just managed to prove that 90%of papers in science are wrong or inadequate, and to kill the entire field of statistics with it!

    • Replies: @Chimela Caesar
  211. utu says:
    @j2

    Ok. No problem.

  212. @j2

    I will dispel your Moon Hoaxer’s fantasies very quicky.
    1) The simple reason Piffer did not use the Jewish sample in the 2015 paper is that back in 2015 there was no Jewish DNA database publicly available. In 2015 he used 1000 Genomes, which does not have a Jewish sample. In the 2019 paper, Piffer included a Jewish sample after the gnomAD database was released in 2017. No conspiracy needed!
    2) The SNPs were identified by Lee et al. on European samples. Piffer did not carry out any GWAS on Ashkenazi or any other population. The PGS was calculated from the results of the Lee et al. GWAS for the populations that he could find on the main public databases. No “new PGS was formed”. Allele frequencies were calculated from the same SNPs for all the populations. There was no adjustement made for any particular population. This is all cheap conspiracy stuff. But now it all makes sense, coming from a moon hoaxer like you.

    • Replies: @j2
  213. j2 says:
    @utu

    This is to support your comment.

    One reason to say that Piffer’s very good correlation is either a coincidence or possibly a result of the selection of a suitable PGS is that country averages even for European countries have changed over the time while genes in those populations have not changed. The reason for the change (like nutrition, education) is irrelevant. It is enough to notice that had Piffer made his study at some other time, his correlation would be worse.

    As an example we can take the year 1980. At that time the Finnish country average was thought to be 97 while the UK average was by definition 100. Had Piffer had GWAS results and plotted his curve, then Finland would not fit to the place where it is now and the correlation would be worse. Ireland is a similar case, Lynn originally claimed that Ireland is a low IQ European country, but this is not so anymore.

    Clearly, Piffer’s good correlation is not a proof of a strong link between the national IQ and PGS, it is a lucky result, or a result of selecting PGS.

    • Agree: utu
    • Replies: @Merculinus
  214. j2 says:
    @Merculinus

    I by mistake addressed an answer to you to utu, comment 210.

    About your understanding of science:

    “A colleague of mine once felt that his work had been plagiarized and complained with the journal’s editor. The editor asked him to submit a paper that would prove that plagiarism had taken place, and this paper would be peer reviewed. The reviewers found some evidence of plagiarism, but since this was not prove beyond reasonable doubt, his allegation wasn’t published.”

    So, your colleague made a baseless claim of that an author had plagiarized. It is a serious charge and naturally, as it was not shown, the editor did not publish such a claim. Had your colleague claimed that this article looks like it is plagiarized, the editor might have said, it does look like that, but we will check it. In the same way, had your colleague stated that this author has forged data, it would be a serious charge, while had he said it looks like the author has forged data, it would be different. Then the author would have thanked him and said, yes, one can get this impression, I must write it in another way.

    “Here we are not asking you to submit to the strict standards of peer-review, we’re asking you to at least do some work and show how exactly Piffer manipulated his data.”

    The table where Piffer has PGS scores for many ethnic groups is problematic. The one for Ashkenazi Jews (2019) refers to a paper where the the PGS was developed in 2018. Some other references are older, therefore they cannot have used the PGS from 2018. This means that either Piffer mixed results using a different PGS, which is manipulation of data, or he has incorrectly given the table in case he recalculated all by using the same PGS. In that case he gives misleading information by adding the references to the table, which is done to add credibility to those results, but if he recalculated them he must not lay the responsibility to other researchers. Both cases are some level of manipulation of data,

    “Piffer obtained the polygenic scores from the supplementary data provided by Lee et al., and his results were very similar using EA_MTAG, CP_MTAG, and EA SNPs, at different significance thresholds, and using the full set, he used weighted and unweighted scores…what else do you want?
    He did not cherry pick SNPs here and there.”

    I have not claimed that he has cherry picked SNP here and there. I have not claimed that he has not calculated some PGS values. It is only a guestion of how the PGS was created. The one created in 2018 from the Wisconsin data was made with a methodological error in case it was to be used for Ashkenazi Jews, as it was. See my comment 210.

    “I invited you to try and reproduce Piffer’s results. Go ahead and compute the Polygenic Scores using Lee’s various GWAS hits, and see if he actually omitted some SNPs to get his fabulous results. Or download his code and data and see if it there are any errors in the code or if the data he used altered Lee et al’s original data. I checked all of this myself and I didn’t find any evidence of data manipulation.”

    You do not understand the problem. I have read carefully Piffer’s earlier papers and there are independent verification that the calculations from that data with that PGS give the results Piffer reports. This I do not need to do. You did it unnecessarily, I was sure they give the results he announced. The problem is elsewhere. What you do not seem to understand is that the selection of the PGS should be done differently, not for fitting data to results that you hope to get, but so that the results you get follow only from the data. The explanation of why Piffer’s PGS was selected in one way or in another way was not explained well enough to remove the doubt of tuning it.

    “Obviously your laziness or lack of skills is preventing you from doing that.”
    Obviously you do not do any ad hominem attacks. I stated that you are very bright, though too stupid to understand the problem in the paper, and that very few supervisors would like a Ph.D. student behaving like you do. You, on the other hand, are very polite and would not claim without basis that somebody lacks skills and is lazy.

    “Piffer provided data and code, now it’s your job to prove that he altered the data.”

    No, it is the author’s job to convince the readers. It is different as with a claim of an author having plagiarized. That must be proven by the one who so claims. But it is the author who claims that his results are correct. He is the one who has to prove that they are correct. He should try to write his paper in such a way that the reader is convinced that the results are correct. Presently, I am not convinced that he has found a real correlation. It looks like his correlation is a result of the selection of the PGS, and he does not give good reasons why just this PGS.

    “Do you want us to believe your allegations simply because you’re old and you have a PhD (assuming you aren’t making this up)?”

    Was it so that you were not making ad hominem attacks?

    “Relying on an academic qualification on an unrelated field plus old age alone aren’t sufficient to prove your point.”

    You have one of my arguments in 210 and another is in a later reply to utu.

    “Your beef against Piffer’s results are based only on these two points:”

    No, my arguments are different. You just do not understand them. I have explained as well as I can. If you cannot still understand, then the problem is on your side.

    “Go back to statistics 101 (if you ever took that course) and study/do homework on measurement errors. Try to measure your height, or your IQ (assuming you have an IQ) three or four times,”

    I have passed several courses on statistics and probability, but was it so that you do not do ad hominem attacks?

    ” and when you get different scores, please don’t come back to this forum claiming that the IQ tests, or the tape were manipulated.”

    Statistical uncertainty is not the reason for growing national IQ values in Finland. Indeed, the military has measured IQ from all conscripted men and that means 20,000-30,000 men. The statistical error is negligible, but there was a trend, first growing, now declining. The genetic basis has not changed.

    “Yes, polygenic scores change as you add more SNPs, and this is measurement error. ”

    No, it is not. It is caused not by a measurement error. You first select the SNPs to your PGS and put them weights. Then you calculate the PGS. If you make a different PGS you get a different result.

    “Again, if you want us to believe your claims, go ahead and show which SNPs Piffer omitted, and how adding those omitted SNPs would worsen his results.”

    You really do not understand what Piffer did. First he used 7 SNPs and calculated a PGS. Then he made a PGS with more SNPs and got a bit different results, for instance, UK went ahead of Finns. Then he again made a different PGS. Still, the SNPs that he uses are only a very small fraction of all SNPs that the human genome has. He selects what SNPs to use. And he selects them from a set of SNPs that are linked to educational achievement, but with explain only a small part of the variance of educational achievement. This selection allows manipulation of the results, whether Piffer has done so it not is irrelevant. It is enough that the method allows it, thus it is not a clear method. I will not make a study of this issue, I do not have time and interest on it. Let us just say, there are SNPs that Piffer used and those that he did not use, and there are different weights to put on them, and by changing the PGS you get different results. I can make one right here. I choose zero SNPs, so the PGS is always zero. Clearly, if I plot IQ against this my PGS, the results differ from those of Piffer. That means, by changing PGS you change the plot, and this is not any statistical measurement error. This is the choice of the quantity to be measured.

    Hope this helps you.

  215. AaronB says:
    @utu

    No Spam for Davide Piffer

    .

    Ouch. That’s just cruel.

    Spam is delicious. The Melanesians are right to hope it arrives from the sky. It’s a luxury high-end food in Korea. A popular food in Japan and Hawaii.

    David Piffer had better hope he can still get his hands on some spam.

    • Replies: @utu
  216. @Merculinus

    Merculinus, you appear to deliberately pretend not to get the point, and rather, try to confound the simple basic logic in Utu’s powerful argument with examples wildly different in principle from the point.

    So, what is the state of research? The SNPs identified by GWAS (the treasure trove), i.e., the “causal” SNPs can predict 3.9% of variance only. To push this prediction to around 10% you need add 1000’s more of SNPs as done by Hsu who used a brute force lasso method and did not waste time on intricate GWAS or by Lee who ended up using circa 1 million of SNPs in his polygenic score.

    The bottom line is that we have no genetic predictor of IQ or education attainment. The heritability gap remains wide open. From GWAS we know of couple thousands of SNPs that on their own can predict only 3.9% of variance, yet we want to predict 50-70% of variance if we want to close the heritability gap. And at this point enters Mr. Chutzpah who happens to find that using the polygenic score of 9 and then of couple thousands of SNPs correlates (r=0.88) with IQs of 26 countries.

    We would need 50-70% variance prediction to close (not partially, but satisfactorily) the heritability gap on IQ and educational attainment.

    You have put up a good fight, but perhaps it is the course of wisdom to know when to, I would not say concede, but to go back to the drawing board to bring forth yet another ambitious work. Criticism and refinement give birth to success.

    • Replies: @res
  217. j2 says:

    Consider the following problem: let there be three populations, A1, A2 and B. Let A1 and A2 have the same average IQ, say 100, and are similar, say European, while B have a lower IQ, say 80 and is say non-European. Assume a GWAS of educational achievement is made based on A1 and B. As people of A1 have a higher average IQ, the GWAS will tend to show SNPs that are more common in A1 and in B as positive to educational achievement, while SNPs that are more common in B will tend to turn out as negative to IQ. Naturally, this is not exactly and only so, because each population also has an IQ distribution, so many SNPs that turn out as positive for IQ will be positive on both subpopulations, similarly for those that are negative. Yet, there will be SNPs that give a positive correlation with IQ that essentially comes only from A1 having a higher IQ than B.

    If we make a PGS based on this study and then use the same PGS to the three populations, A1, A2 and B, the expected order is A1 gets the highers score, then A2, as it is somewhat similar to A1 genetically, yet not the same as A1. B will get the lowest score.

    However, if we make a PGS from a GWAS in A2 and B, and apply this PGS to the three populations, the order will be A2, A1, B.

    This seems to be what Dunkel et al did and it is methodologically wrong to do this.

    If we make the GWAS oin similar populations A1 and A2 and apply it to the three populations A1, A2, B, then there comes another error: B gets a low score because it is not similar to A1 and A2, not necessarily at all because it is of lower IQ. This is the problem that there may be special SNPs in B.

    What should be done is to measure similar populations together and to measure each different population individually to see what SNPs there should be and what weights they should have. Then the results can be combined.

    What should have been done by Dunkel et al is to make a separate GWAS only on Ashkenazi Jews, but for this 53 is too small, and then find SNPs from it and then combine them to the PGS. As this was not done, the study has a serious methodological error. As Piffer used these results, his paper also has a serious problem.

    I consider both papers as broken to pieces beyond repair.

    • Replies: @res
    , @Merculinus
  218. j2 says:
    @Merculinus

    “But now it all makes sense, coming from a moon hoaxer like you.”

    This kind of mockery is the limit of your scientific abilities. You have not produced a single argument of any value. You, like trolls of Unz in general, are just for dirt throwing, nothing else.

  219. res says:
    @Chimela Caesar

    We would need 50-70% variance prediction to close (not partially, but satisfactorily) the heritability gap on IQ and educational attainment.

    Individual prediction is very different from predicting group average. Piffer is doing the latter as I explained at length in comment 148. Perhaps it wold be more clear if I explicitly said “prediction of group averages” rather than “group prediction”?

    Again, the important question is:

    Why does Piffer get such consistently (multiple replications) good results given all of the issues with the methodology?

    P.S. I really wish I understood some of the motivations here better. When I see consistent misunderstanding of rather simple simple points (like individual vs. group here) I wonder what is going on.

    • Replies: @Chimela Caesar
  220. res says:
    @j2

    Piffer uses Dunkel’s results for only a small (confirmatory) portion of his work. Invalidating Dunkel would by no means invalidate Piffer’s paper.

    But let’s discuss Dunkel’s work then. Emil’s FAQ page for the paper (which you linked above) seems like a good reference to use.
    https://emilkirkegaard.dk/en/?p=7680

    He directly addresses your concern in “But how do we know the polygenic scores are valid in Ashkenazim?”

    While we are not aware of any direct validation study, it is known from plant, animal, human and simulation research that polygenic score validity declines as a function of Fst between the training population and the target population (Scutari et al 2016). The Fst difference between Ashkenazim and Central and Northern Europeans is tiny — about 0.06 to 0.08 according to Bray et al 2010 — so we don’t expect any serious decrease in validity.

    You might also look at the following FAQs:
    – What didn’t you use population stratification controls?
    – But why didn’t you use more population stratification controls? Look at this random study which did that
    – Only pseudoscientists don’t use population stratification controls!

    A large scale Jewish sequencing project (akin to the UKBB) would definitely help improve the Jewish PGS results though. I have seen relatively few Jewish GWAS though, and they tend to have sample sizes far to small to resolve many EA or IQ SNPs.
    https://www.ncbi.nlm.nih.gov/pubmed/21812969
    https://www.ncbi.nlm.nih.gov/pubmed/26198764
    https://journals.plos.org/plosgenetics/article?id=10.1371/journal.pgen.1002559

    And again, the EDU PGS gives good seeming results for Jews in both Dunkel’s WLS data and Pallesen’s HRS replication at http://rpubs.com/Jonatan/jewish_pgs

    jp7 and I (comments 197 and 192 respectively) have responded to your objections regarding Pallesen’s work. Hopefully those comments meet your “single argument of any value” criterion from comment 221.

    You definitely enjoy the cut and thrust of academic debate. I can see how you would succeed at a university. I just wish you would produce more arguments of value yourself and fewer/less ad hominems and rhetoric.

    P.S. There is a great deal of “isolated demands for rigor” occurring in this thread. This is a useful look at that tactic:
    https://slatestarcodex.com/2014/08/14/beware-isolated-demands-for-rigor/
    I can at least understand it in j2’s case since he comes from a hard science background where the standards of proof tend to be much higher.

    • Replies: @j2
    , @j2
  221. j2 says:
    @res

    About my participation to this discussion. James Thomson in the article wants the commenters to break Piffer’s paper into pieces. As that was the request, I did so. I do not basically have anything against Piffer and his papers, but if the task is to break it to pieces, it takes me maybe a day or two to find an error that can be considered fatal. Whether you consider it fatal is your choice. For me, as a part of the task, it is enough to find an error that on a stricter field would be considered fatal. I found such an error and that is the end of this exercise for me. Just an intellectual challenge, break a paper of a nice guy into pieces because some post author asks you to do it.

    About your answer:

    In Dunkan et al Fst is not a problem, but it is in Piffer. The Fst distance between Finns and Ashkenazi Jews is about 0.013-0.023. Piffer’s paper compares Ashkenazi also with Africans and Asians.

    But regardless of Fst, you have the problem with the populations A1, A2 and B. Your reference does not consider this problem, you get a bias because of the way you select the PGS. Your reference thinks of a different usage for the PGS. That is, your PGS is still good for Ashkenazi, but it is not good for comparing Ashkenazi with other people. As you have A1=Ashkenazi, B=Christians of Wisconsin, you automatically end to a problem when comparing Ashkenazi with another European ethnic group.

    There definitely should be a study that includes Separdic/Mizraim Jews as they share genes with Ashkenazi and only in that way we could know if the used PGS reflects Ashkenazi origins too much and not enough educational achievements.

    About ad hominem attacks, they were started by Merculinius. When there is such a trolling person in the discussion it is difficult to keep it in a scientific level. Your comments were quite fine and I am sorry that I did not make a clear difference between you and Merculinius. He started with giving me a conspiracy theoretician prize. That is not a good start at all.

    About you being incredibly stupid, which I did wrote and hurt your feelings, I did not naturally mean stupid in the normal sense, you sound very intelligent. It was also not of you not having a doctorate, that could not matter none to me.

    What I meant when writing it is that many such clever people like you are incredibly stupid in the sense that they have not yet realized that much of what is claimed to be true is lies. This science that you think is honest work (and I do not mean especially Piffer, I mean much a wider range, much of what intelligent people normally believe) is not true. And because of that one cannot trust researchers as much as we once believed we could, and this does include Piffer. There are people with agendas working in what looks like science. Such naive people who do not know this should learn life more and take a red pill, get some wisdom, and notice that things are not as they should be in science. They do cheat. it is even that (((they))) do cheat, believe me or not.

    • Replies: @res
  222. More on Piffer soon.

  223. @j2

    Again, you are cherry picking the countries that have changed IQ in a direction that would weaken the correlation. You have proven absolutely nothing. You omitted those that changed their values so as to make the correlation stronger.You should change all the IQs and set them to some other date and see if the correlation holds. Your logic is again flawed here. You can’t have a PhD and if you do, you got it so long ago that it expired.

  224. @j2

    You keep getting things wrong. The reference to Dunkel et al in Piffer’s paper refers to their estimate of phenotypic IQ, as the other references also refer to estimated IQ (E.G. Dutton and Kierkegaard for Finns) The PGS was calculated independently from Dunkel et al and it is the same that was used for the other populations.

  225. j2 says:
    @res

    Maybe I try to make up the unintended ad hominem attacks. I did not make any such attacks, but you experienced my words in such a way. So, here I have an argument. I hope it makes you open your eyes and see why I did not like you to defend a paper that very possibly is intentional fraud. I hate to see nice people defend fraud thinking they defend good science against unfair attacks of a crazy conspiracy theoretician.

    You found an explanation what they have prepared to answer questions. One that you found was that if Fst difference is small then they can make the PGS the way they did. Is this so? No. They refer to animal and plant studies, but if Fst difference is small, then these populations are similar. Similar populations can be treated together. So, are Ashkenazi Jews similar to Christian Americans in the respect to the measure that they calculate, that is, IQ? No, they claim that there is a 10 point difference. How can these populations be similar with respect to the only relevant measure in this study if the researchers found out and state that they are not at all similar but differe by 10 points.

    Thus, this reference to Fst is just exactly bullshitting the people who ask these questions. It is not that they do not understand it. Naturally they do. Anybody who a bit thinks of it must notice, no, these populations are not similar with respect to IQ. So, let’s use some paper that talks about plants and animals, those are similar if Fst difference is small, so this way we will settle this question.

    And think about one IQ study which found that US Jewish IQ is 110 and the White Christian average is 106. Putting White average to 100 gives the Jewish IQ as 104, but it is everywhere announced that the study found it to be 110.

    • Replies: @res
  226. res says:
    @j2

    Thanks for your more civil reply. Point taken about the blog post inviting criticism.

    Just an intellectual challenge, break a paper of a nice guy into pieces because some post author asks you to do it.

    That makes sense. Especially for a retired academic. Much of my objection has been to a sense that you were not responding to solid responses to your original criticism. When someone receives both solid criticism and insults and then chooses to focus on the insults (which I think is an accurate description of many of your comments in this thread) I question how much they really want to have a legitimate discussion of the paper’s merits.

    In Dunkan et al Fst is not a problem, but it is in Piffer. The Fst distance between Finns and Ashkenazi Jews is about 0.013-0.023. Piffer’s paper compares Ashkenazi also with Africans and Asians.

    The important Fst number is not the distance between two extreme populations. It is the difference between each population and the population used in the original GWAS (this is what affects PGS accuracy). Here that would be the UKBB so British of European origin. I suspect Finns and AJ are both closer to the UKBB population than they are to each other.

    Supplementary Figure 2 of
    https://media.nature.com/original/nature-assets/tpj/journal/v18/n1/extref/tpj201677x1.pdf
    has a Fst matrix graphic for the 1000 Genomes populations.
    The continent level races have Fsts more like 0.1 than 0.01.

    Does anyone have an Fst matrix for the gnomAD populations? Where did you (j2) get your Finns-AJ Fst numbers?

    Regarding

    What I meant when writing it is that many such clever people like you are incredibly stupid in the sense that they have not yet realized that much of what is claimed to be true is lies.

    I agree with this concern, but it surprises me to see it raised in this context. In my experience there are far more lies coming from blank slatists when it comes to the topics of IQ, genetics, and their relationship.

    Stepping back from the Piffer debate for a moment, how likely do you think it is that the observed differences in country IQ have some degree of genetic cause? Any thoughts on the magnitude of that effect? IMHO the people who deny there is any genetic effect are the real liars in this conversation.

    To my mind the genetic basis of IQ (hopefully leading to more physiological understanding someday) and the relationship of that to both group and individual differences in IQ (and thereby to other metrics, see Rindermann’s work) are the central questions.

    I think Piffer’s work is some of the best evidence we have that between group IQ differences have a significant genetic basis. To my mind that makes it extremely valuable (and also much hated, mostly by liars though not ALL critics are liars about genetics and IQ). There are certainly methodological issues (mostly due to lack of data, large sample studies like the UKBB replicated on other groups would make this work much more compelling), but I have yet to see a criticism that calls into question the conclusion that genetics has a significant effect on between country IQs.

    P.S. Calling me stupid does not actually hurt my feelings (though it is annoying). It is so obviously wrong that it mostly makes me laugh and question the competence of the source. I do feel a need to speak up in my own defense though. Most people don’t respect someone who lets others abuse him. Even (perhaps especially?) when the criticism is obviously ridiculous.

  227. res says:
    @j2

    You found an explanation what they have prepared to answer questions.

    You were the one who advocated that authors defend their work. I went to what I think is the best defense made of the Dunkel paper by one of the authors. Seems like the sensible thing to do.

    Beyond that, the appropriate thing to do is address the argument made there. Which, to your credit, is what you did next.

    You found an explanation what they have prepared to answer questions. One that you found was that if Fst difference is small then they can make the PGS the way they did. Is this so? No. They refer to animal and plant studies, but if Fst difference is small, then these populations are similar. Similar populations can be treated together. So, are Ashkenazi Jews similar to Christian Americans in the respect to the measure that they calculate, that is, IQ? No, they claim that there is a 10 point difference. How can these populations be similar with respect to the only relevant measure in this study if the researchers found out and state that they are not at all similar but differe by 10 points.

    I am not following your reasoning here. I followed the Scutari et al 2016 link:
    https://journals.plos.org/plosgenetics/article?id=10.1371/journal.pgen.1006288
    I did not read the entire paper, but from a quick look they are looking at populations under artificial selection. Given that, I would expect the phenotypic differences in the trait in question between groups (which is being selected for!) would be even larger relative to Fst than we see in IQ between Christians and Ashkenazi Jews. Thus your concern about the latter being “not similar” relative to the former is if anything backwards.

    Perhaps you could elaborate your argument?

    One thing that is not at all clear to me is if/how the difference between predicting individuals (Scutari) and predicting group averages (Piffer, Dunkel) matters here.

    And think about one IQ study which found that US Jewish IQ is 110 and the White Christian average is 106. Putting White average to 100 gives the Jewish IQ as 104, but it is everywhere announced that the study found it to be 110.

    Agreed that games like this are played in many places. We all need to be alert for them. That’s one of the reasons I ask for and follow references so frequently.

    • Replies: @j2
    , @j2
  228. @res

    This is the unstated inference.

    When you have a dataset system, its validity and quality would or should depend on the validity and quality of its data points.

    If the data points produced by a genomic algorithm are not healthy enough due to low variance prediction, what would you think, logically, of the system built off these points?

    We would need 50-70% variance prediction to close (not partially, but satisfactorily) the heritability gap on IQ and educational attainment.

    • Replies: @res
  229. Some replies and clarifications from Piffer tomorrow, in answer to a reviewer whom I had contacted before posting up Piffer’s work, and also as a way of answering some of the points made here, though I can see that some of them have already been answered.

    • Replies: @j2
  230. res says:
    @Chimela Caesar

    If the data points produced by a genomic algorithm are not healthy enough due to low variance prediction, what would you think, logically, of the system built off these points?

    It depends. Think of it as a signal processing problem. Statistical averaging can be a powerful technique for extracting useful information from a noisy signal.

    When you have a dataset system, its validity and quality would or should depend on the validity and quality of its data points.

    I disagree. Its validity and quality depends on if and how well it works. And Piffer’s method seems to both work and reproduce consistently so far.

    • Replies: @Chimela Caesar
  231. AaronB says:
    @utu

    Yes!

    That’s the good stuff!

    If pfiiffer retracts his silly statements we shall let him have some.

  232. j2 says:
    @res

    “Perhaps you could elaborate your argument?”

    Consider this my comment of three populations A1, A2 and B where A1 and A2 are rather similar and both have the same IQ, and P has a lower IQ. If you make GWAS on IQ related thing on A1+B where these populations are put together, this your GWAS will react not only to SNPs that in each population A1 and B correlate with IQ, but you will also get SNPs that correlate merely with belonging to each group.

    In the 2019 study of Ashkenazi Jews Jews and Christians were grouped together. Indeed, with 53 Jews they could not make a GWAS on Jews alone, so they were grouped together, even though it was known that in this sample of Jews born in 1957 there was a 8-10 point IQ difference and the GWAS was to measure IQ. Thus, they knew exactly well that they have the problem with A1,B populations.

    What would have been a normal thing to do is to take PGS that was developed from similar European populations. That PGS is likely to be rather good, as it is developed from populations that do not much differ in IQ, therefore it should not much reflect ethnic group . (There is some small concern between North and South Europe, they actually should be measured separately and then the PGS carefully joined.)

    Dunkel et al could have taken such a ready PGS and concluded that Ashkenazi Jews, as genetically quite similar to Italians, do not differ very much from European populations, small Fst distance to Italians, so it is reasonable to use this PGS to Ashkenazi. I would have agreed. It is used by Piffer to East Asians and Africans. But they did not do so, they went on to make a new PGS from the Wisconsin study. That is extremely odd and extremely wrong. Someone among the authors must have realized that there is the A1,B problem. Then using this measure to A2, a European population, you get the order A1, A2, B. That is, you make intentionally a PGS that correlates not only with educational achievement but also with Ashkenazi origins. We can see that the PGS does this from the plots of the HRS study: cognition (a proxy for IQ) differs less than the two PGS scores, thus the PGS scores correlate with ethnic origins. We can less clearly see it from the data points in Dunkel’s plot Plot of PGS vs IQ. It does look like the Ashkenazi points are more spread in the horizontal direction than in vertical. It indicates that the PGS correlates with Ashkenazi origins.

    And instead of explaining how you avoided this obvious problem, the authors try to explain it off with a phony explanation that the Fst distance is small.

    How I get the Fst distance of Finns and Ashkenazi Jews. You find the rather large Fst distance between Finns and Italians and the very small Fst distance between Italians and Ashkenazi Jews and then notice that these are not added linearly but at most orthogonally, but as the second Fst distance is so small, the result is the Fst distance between Italians and Finns.

    Piffer maybe used the 2018 Wisconsin PGS to all European populations. It does not look like that as he gave dates of studies in his table. The values he announces should be from those studies and if so, the values are from different PGS and the Ashkenazi score is from a PGS that correlates with Ashkenazi origins due to the incorrect way it was created. Even if Piffer did correctly and recalculated all samples with the Ashkenazi PGS, he could not remove the problem with this PGS: it still correlates with Ashkenazi origins. It will be that any European A2 will typically fall under A1.

    About your other questions. I read earlier Piffer’s papers carefully long ago, this one I was not interested in as it seemed to have a problem with the Ashkenazi score and Thomson wanted the paper broken. But the earlier I read. It is difficult to point out what is the reason. I do believe there is a genetic basis to IQ differences between the main races.

    One concern: The PGS was done in Europe, most probably rather well. The small genetic differences between North and South Europe may be reflected in the PGS as indicating ethnic origins of the North or the South. As the researchers of this field may? not understand this problem (which would be amazing from researchers, but OK, it is a soft field), then possibly there could be this bias. Or they do understand but have an agenda, this is also possible.

    For Eastern Asians the PGS created for Europeans gives higher scores than for Europeans. This is not necessarily suspicious. The PGS may be correctly formed and East Asians are not so far from Europeans. They may have the same SNPs that determine IQ. So, this I do not immediately oppose.

    With Finns I have some problem. Piffer’s scores are of course correctly calculated from the data and the data for Finns is large and representative. Yet, some foreigners, Ed Dutton, D Piffer and some IQ people, claim that Finns have a higher IQ than other Europeans. The East Asian admixture is from 4% to 15%. It is not that large. Most Finns attribute PISA to the school. I attribute it to the school, the language and personality. (introverts are more intelligent than extroverts as they think and read more, not because of being genetically smarter).

    About Africans, I think that they are so far in Fst that they may have different SNPs. That should yet be studied. It is so that we know there is an IQ difference between Africans and Europeans, and we know that there is a genetic difference between them. Thus, a PGS made for Europeans would usually give a smaller score to non-Europeans as they are different. (Notice the comment of East Asians, but they are not so different from us). And this difference would not have any relation to IQ, but there would also be an unrelated IQ difference and they you correlate them and use one as a causal reason of the other. That is a problem. Piffer’s plot is not verified from the part outside Europeans and East Asians.

    There are indeed liars, who claim that there are no IQ differences. But the liars do not say exactly that. They say: there are no IQ differences between non-Jewish populations, but there is 15-20 point difference between any non-Jewish population and Jews. Then, if you press, the claim will modify to, only Ashkenazi Jews, and only 10 points. This claim is necessary for explaining the high representation of Jews. Though, it does not explain the high representation of Jews as statistically they are heavily overrepresented even if we assume an IQ of 110. The actual IQ difference today is 103.5, the best measure for young adults of today. For my age American Jews, the difference was bigger, for both selection and environmental reasons. They need this overrepresentation in science and elsewhere as a part of control of finance, media and science, which is the set used in this case.

    • Replies: @Merculinus
    , @res
  233. @j2

    “The values he announces should be from those studies and if so, the values are from different PGS and the Ashkenazi score is from a PGS that correlates with Ashkenazi origins due to the incorrect way it was created”.
    I am sorry but you got this entirely wrong and this caused a big misunderstanding.Perhaps you have a deep mistrust of science, I don’t want to think that you’re a schizo. You also admitted to having only skimmed through Piffer’s last paper, so you’re not qualified to comment on it.
    Please stop spreading this lie. The values are not from different PGS. Piffer did not use the PGS from the Wisconsin study for the Ashkenazi. Nowhere in the study does he mention that the PGS were re-calculated separately using the Wisconsin study. You misread the reference in table 5, where at the IQ column the IQ estimates are reported. Lynn and Vanhanen are cited alongside Dunkel et al. Did Lynn and Vanhanen conduct any GWAS? No. These citations refer to the estimates for average measured IQ. NOWHERE DOES PIFFER MENTION THE WISCONSIN STUDY. Your arrogance goes so far as to make up stuff and bring down an entire paper with it by misinterpreting both the methods and the data that were used. This is not acceptable.
    He used the Lee et al. 2018 GWAS done on Europeans and calculated PGS on Ashkenazi and all the other groups from the same set of GWAS hits. “Among the 2416 GWAS significant SNPs (Lee et al., 2018) [7] 2404 SNPs were found in the gnomAD dataset. ” That’s what he says. This work bears no relation whatsoever to Dunkel or Pallesen’s study.
    The references in the table are for the estimates of phenotypic population IQ, not for the PGS.
    The following claim is also naive and reveals your poor judgment throughout: ” Thus, a PGS made for Europeans would usually give a smaller score to non-Europeans as they are different”.
    This is not necessarily so. Population-specific variants can have both IQ enhancing or depressing effect. Africans would get higher score than with the present GWAS only if there was an overrepresentation of positive effect alleles for population specific variants.
    Piffer cites a GWAS carried out on Peruvians where they found an height-decreasing allele (an inch for every allele copy). You naively assume that an African GWAS will only find IQ boosting mutations, yet the opposite can happen.
    How come the PGS done for Europeans gave a higher score to East Asians? And how come the height GWAS gave a higher score to Africans and a really low score to East Asians (Yet following your logic, Africans should always get lower scores than Asians because they’re more different from Europeans).
    You can see that you’re wrong all over the place.

    • Replies: @j2
  234. j2 says:
    @res

    You were irritated by me not taking your explanations into account, but what would you do yourself in a following situation:
    You talk with a religious group member about some supremacy claim this religious group makes about themselves and you know it is false.
    Then walks in an intelligent outside, who takes a book that this religious group prints for their children entitled: “What do I answer if the non-believers claim that our holy books are not correct?”
    and then this outsider starts reading aloud the correct answers.
    Would you say, where the hell did you pop up from and cannot you see that those are all lies? Or would you patiently start teaching this outsider what exactly is false with those arguments while your original opponent stands by and smiles being satisfied by your unexpected support?

    • Replies: @res
  235. j2 says:
    @James Thompson

    I have never been especially interested on GWAS for educational achievement, but after a very brief look I suggest that the researchers of the field investigate if:
    – when doing a GWAS study that measures some property, they take care not mixing populations that differ in this property because it results into correlation between the measured property and belinging to a specific population.
    – check if this kind of error has been made in European data. Northern and Southern Europeans slightly differ in IQ, but they also differ in genes and the PGS may partially reflect belonging to some ethic group, not only educational achievements.
    – check if personality traits, notably extroversion/introversion may explain the North South IQ gradient. In some cultures, those with books and a habit of reading them, introversion is likely to increase IQ, while in other cultures it lowers (like country side people without any information are less intelligent than city people, so in this case information comes largely from contact with people)

    I hope my comments helped you, for my part this is enough of this particular topic.

    • Replies: @res
  236. @res

    When you have a dataset system, its validity and quality would or should depend on the validity and quality of its data points.

    I disagree. Its validity and quality depends on if and how well it works. And Piffer’s method seems to both work and reproduce consistently so far.

    See the context of the system “validity”. Consider the function f a system:

    f(x) = x2, valid for all real values of x

    • Replies: @res
  237. res says:
    @j2

    In the 2019 study of Ashkenazi Jews Jews and Christians were grouped together. Indeed, with 53 Jews they could not make a GWAS on Jews alone, so they were grouped together

    You are demonstrating a fundamental misunderstanding here. There was no GWAS done in the 2019 study of Ashkenazi Jews Jews and Christians.

    Here is Piffer’s process as I understand it. Others please correct me if in error.

    1. Gather a set of SNPs from an existing GWAS. As I understand it, the recent studies are using Lee et al. 2018 based on the UKBB. That particular study provided two lists of SNPs (all and “causal”). Piffer looked at both.
    2. Create one or more PGS from the GWAS SNPs. This typically involves some form of weighting and or significance thresholding. Those are done algorithmically (e.g. p-value weighting/threshold) which lessens the opportunity for cherry picking. Especially when “typical” threshold values are used along with no weighting.
    3. Use those PGS to evaluate populations from other studies which provide both genetic and phenotypic data. Here the WLS and HRS.

    So Piffer was using the highest powered EA GWAS we currently have. The problem you describe in the excerpt above does not exist. The issue of applying that GWAS to other populations is present (small for cases like Ashkenazi Jews, larger for Africans and Asians) and has been discussed already.

    I hope that was clear. I skimmed the rest of your comment (see PS), but am mostly holding off responding until we get this very fundamental point sorted out.

    I believe it is accurate to say that a GWAS is concerned with finding the SNPs (or other genetic info) corresponding to a trait (hence association). It would be more accurate to characterize Piffer’s work as the application of results of a GWAS.
    https://en.wikipedia.org/wiki/Genome-wide_association_study

    P.S. Regarding your last paragraph, there is a whole taxonomy of liars in this area. Some of the common forms.
    1. There are no average IQ differences between groups. Pretty much discredited, but still seen occasionally.
    2. There are average IQ differences between groups, but genetics play NO part in that. This seems most common in mainstream science right now. It is also the one in the process of being invalidated by recent work–like Piffer’s. I tend to focus on it.
    2a. I get the idea we are starting to move on to a variant of 2 where the genetics differences exist, but are too small to be meaningful. As some have noted, this is a slippery slope.
    3. Various forms of group supremacist wrangling. The Jewish example seems to be the most common. I tend to have a foot in both camps in that I think Jewish success is a mix of higher average ability and ethnic nepotism (etc.) rather than either alone.

    On to the more recent forms like some of what we see from Harden and Turkheimer. I would not call that outright lying though, more like dissembling IMO. 2a above often falls into this category depending on which exact assertions are made.

  238. res says:
    @j2

    That is a useful analogy. It helps explain how you are seeing this and what you are reacting to.

    I see the Jewish aspect of this as a sideline to Piffer’s work (this is less true of the links to Dunkel and Pallesen, though I have no sense of their proclivities. Emil (one of three authors on the Dunkel paper, and author of the web page in question) strikes me as following the data so I tend to give him the benefit of the doubt) AFAICT Piffer only brought in the gnomAD data they used because it made a good confirmatory test of his work. He has been focused on the 1000 Genomes data previously, and that still makes up the bulk of his current paper.

    So rather than walking in on a religious group ,we are walking in on a conversation about something else (Piffer) where an academic paper produced by someone else (of unknown affiliation, so should treat with the usual cautions and standards of proof/debate) has been brought up as relevant. We can also assume (I hope) an audience of good faith willing to consider the arguments on their merits.

    I have my own share of hot buttons, and this is a controversial topic, so I am sympathetic. But I do think we should try discussing Piffer’s (and related work) on its merits rather than dragging ethnic biases and animosity into this. (I need to keep this in mind as well)

    P.S. Thank you for engaging with me in a constructive fashion. I enjoy conversations like this even if (perhaps even more if ; ) there is occasional disagreement. Smart and knowledgeable critics who strongly but conscientiously argue the other side are invaluable for vetting work. I think that is the spirit in which Dr. Thompson’s original request for criticism was made.

  239. res says:
    @Chimela Caesar

    See the contexts of “Don’t let the perfect be the enemy of the good,” and “models.”

    And I am curious about the source of your definition of system validity. These pages seem relevant:
    https://en.wikipedia.org/wiki/Validity_(statistics)
    http://www.promodel.com/onlinehelp/ProModel/80/C-03%20-%20Model%20Validation.htm

    • Replies: @Chimela Caesar
  240. Yes that is a good description of the process used by Piffer. J2 got it completely wrong: there was no GWAS carried out on Jews, and Piffer did not use Dunkel et al’s WLS’s results. Actually Piffer’s Jewish sample is probably not even part of the WLS (at least not explicitly so). It was obtained from gnomAD, which strictly provides aggregate data only and does not say the source of the samples (for privacy reasons I suppose). The SNPs for the new PGS in Piffer’s 2019 paper were gathered from Lee et al’s summary statistics, which included only European descent populations.
    J2 here reveals his complete lack of familiarity with GWAS, as he seems to think that a few hundred individuals are enough. Even 100,000 is considered a small sample for a GWAS nowadays.
    J2 didn’t read Piffer’s paper so it is not surprising that he got it all wrong. Piffer’s paper never mentions the WLS and cites Dunkel only with reference to their estimate of phenotypic IQ, as it cites Lynn and Vanhanen.
    It is not acceptable that someone attacks a paper after only skimming through it.I’d say we can move on and ignore j2’s comments until he familiarizes himself with GWAS and actually reads Piffer’s work.

    This is Piffer’s response in the next blog post:
    ” I would like to add a note here to address some criticism from readers on this forum. The reference to Dunkel et al. could mislead some readers. It refers to their estimate of phenotypic IQ (110) and in no way to their method of PGS calculation. Their estimate is drawn from an unsystematic review of other studies and can be considered a best guess. The PGS for Jews was calculated just like for the other populations, in a manner totally independent from Dunkel et al’s work. I am sorry if some readers might have been misled and I wish I had made this clearer in the paper.”

  241. j2 says:
    @Merculinus

    “Perhaps you have a deep mistrust of science, I don’t want to think that you’re a schizo. ”

    Was it so that you do not do ad hominen attacks?

    Looking at Dunkel’s paper
    https://www.researchgate.net/publication/330601752_Polygenic_Scores_Mediate_the_Jewish_Phenotypic_Advantage_in_Educational_Attainment_and_Cognitive_Ability_Compared_With_Catholics_and_Lutherans

    we find this place:
    U.S. years-of-schooling equivalents (Lee et al., 2018). The polygenic score for educational attainment used in this analysis (PGS_EA3_MTAG)

    This seems to indicate that Lee et al, 2018 is the PGS with the name PGS_EA3_MTAG

    Looking for PGS_EA3_MTAG we find only these documents
    http://jsmp.dk/files/wls_data.html
    which is from the Wisconsin study.

    https://www.cambridge.org/core/journals/twin-research-and-human-genetics/article/evidence-for-the-scarrrowe-effect-on-genetic-expressivity-in-a-large-us-sample/A428EBF64787D00B3A37DDA61D41D967
    also referring to the Wisconsin study

    https://www.biorxiv.org/content/10.1101/429860v1.full
    also referring to the Wisconsin study

    https://rstudio-pubs-static.s3.amazonaws.com/451760_13482aaa5ffa4c5d9ec06002645935f2.html
    also referring to the Wisconsin study

    This seems to indicate that the PGS_EA3_MTAG is a PGS formed for the Wisconsin study and as it is the PGS of Lee et al 2018, then the PGS (Lee et al, 2018) is the PGS from the Wisconsin study, and that was used by Piffer.

    Naturally google is not 100% certain, but all references to PGS_EA3_MTAG are to the Wisconsin study and Dunkel et al explicitly identify Lee et al 2018 by PGS_EA3_MTAG.

    Assuming that Dunkel et al, 2019, calls PGS_EA3_MTAG incorrectly Lee et al, 2018, then it is their mistake. I cannot notice such an error, I must trust them that they know what PGS they use.

    Anyway, this is the way I concluded that PGS (Lee et al, 2018) is in fact PGS_EA3_MTAG which was formed for the WSL study. If so, it was naturally a methodological error. If not, then Dunkel et al write incorrectly. Or possibly Lee et al made another PGS in 2018 and these two papers, Piffer and Dunkel do not care to mention which of the two different PGS (Lee et al, 2018) they use. That is also a mistake.

    Obviously you did not try to check what the PGS was and what is PGS_EA3_MTAG.

    • Replies: @j2
  242. j2 says:
    @j2

    “Anyway, this is the way I concluded that PGS (Lee et al, 2018) is in fact PGS_EA3_MTAG which was formed for the WSL study. ”

    It is naturally possible that Lee et al, 2018, did not use data from the Wisconsin study and the Wisconsin study just happens to be the only study where this PGS was used, but as the Wisconsin study probably was going on or at least prepared in 2019, there very possibly was contact with Lee et al and Dunkin et al, which means that the PGS was probably tried with the Wisconsin data. That alone makes it suspect. Dunkin et al could have used some other PGS and they should have got data from Israel form Mizraim/Sephardic Jews. As the study is, it raises questions.

    For Piffer to refer to Dunkin et al sample of 53 Ashkenazi Jews born in 1957 for an IQ score would have been incorrect. That sample is too small and that cohort is too old to be reliable.

    In my comments I was mainly interested in Dunkin et al and I read that paper. Piffer in general does not interest me at all. Even the case that he took an IQ measure for Ashkenazi Jews from Dunkin et al is a sufficient reason to discard Piffer.

  243. res says:
    @j2

    – when doing a GWAS study that measures some property, they take care not mixing populations that differ in this property because it results into correlation between the measured property and belinging to a specific population.

    GWAS researchers are well aware of this problem and go to fairly extreme lengths to minimize it. Typically all non-Europeans are removed from studies at the outset. Then correction for the principal components of the genetic population structure is done. The UKBB provides principal components data for (I think) the first 20 PCs. If you are interested in learning more about this, “population structure” and “principal components” are good search terms.

    The GWAS in question is Lee et al. 2018. Here is the paper and abstract.
    https://www.nature.com/articles/s41588-018-0147-3

    Here we conducted a large-scale genetic association analysis of educational attainment in a sample of approximately 1.1 million individuals and identify 1,271 independent genome-wide-significant SNPs. For the SNPs taken together, we found evidence of heterogeneous effects across environments. The SNPs implicate genes involved in brain-development processes and neuron-to-neuron communication. In a separate analysis of the X chromosome, we identify 10 independent genome-wide-significant SNPs and estimate a SNP heritability of around 0.3% in both men and women, consistent with partial dosage compensation. A joint (multi-phenotype) analysis of educational attainment and three related cognitive phenotypes generates polygenic scores that explain 11–13% of the variance in educational attainment and 7–10% of the variance in cognitive performance. This prediction accuracy substantially increases the utility of polygenic scores as tools in research.

    The Supplementary Material describes their analysis (note the correction for first 20 PCs):

    1.5. Association Analyses
    Cohorts were asked to estimate this regression equation for each measured SNP:
    𝐸𝑑𝑢𝑌𝑒𝑎𝑟𝑠 = 𝛽0 + 𝛽1 𝑆𝑁𝑃 + 𝑷𝑪 𝜸 + 𝑩 𝜶 + 𝑿 + 𝜖, (1.1)
    where SNP is the allele dose of the SNP; 𝑷𝑪 is a vector of the first ten principal components of the variance-covariance matrix of the genotypic data, estimated after the removal of genetic outliers (we instead used twenty principal components in UKB analyses); 𝑩 is a vector of standardized controls, including a third-order polynomial in year of birth, an indicator for being female, and their interactions; and 𝑿 is a vector of study-specific controls. Cohort analysts were asked to impose a number of standard subject-level filters prior to running the analyses. These include: (i) each subject’s EduYears was measured at an age of at least 30, (ii) each subject passed the cohort’s quality control, which always include the removal of genetic outliers and individuals with poor genotyping rates, and (iii) each subject is of European ancestry.

    Here is how they used their PGS for prediction out of sample (note the correction for first 10 PCs):

    Polygenic predictors derived from earlier GWAS of EduYears have proven to be a valuable tool for researchers, especially in the social sciences6,7. We constructed polygenic scores for individuals of European ancestry in two prediction cohorts: the National Longitudinal Study of Adolescent to Adult Health (Add Health, n=4,775), a representative sample of American adolescents; and the Health and Retirement Study (HRS, n=8,609), a representative sample of Americans over the age of 50. We measure prediction accuracy by the ‘incremental R2’ statistic: the gain in the coefficient of determination (R2) when the score is added as a covariate to a regression of the phenotype on a set of baseline controls (sex, birth year, their interaction and 10 principal components of the genetic relatedness matrix).

    P.S. IMO they go to such extreme lengths they are sacrificing significant signal as well. It would be interesting to know how much variance was explained by correcting for the first ~10-20 PCs of population structure. But for some reason that information never seems to be published.

  244. j2 says:

    Thanks res, you have just verified what I said.

    This is from:
    Dunkel et al, 2018, Polygenic Scores Mediate the Jewish Phenotypic Advantage in Educational Attainment and Cognitive Ability Compared With Catholics and Lutherans. Available from: https://www.researchgate.net/publication
    “in addition to self-reported mathematical ability and highest mathematics class successfully
    completed. This multivariate PGS was selected because it likely captures the largest degree of
    shared (i.e., GCA-like) genetic variance common to these cognitive phenotypes.”

    This is text copied from your comment:
    “We constructed polygenic scores for individuals of European ancestry in two prediction cohorts: the National Longitudinal Study of Adolescent to Adult Health (Add Health, n=4,775), a representative sample of American adolescents; and the Health and Retirement Study (HRS, n=8,609), a representative sample of Americans over the age of 50.”

    From these two texts notice that 1) in the HSR study there were 153 Ashkenazi Jews and a larger set of other White Americans, not Europeans of Europe. 2) The criteria for educational achievement included self-reported mathematical ability in addition to objectively measured academic achievements of IQ etc.

    Notice what this means: we have in the GWAS a group consisting of Ashkenazi Jews and other American White people (with a major subset called Christians). This is exactly as I wrote to you, it is A1+B populations measured together where A1 has higher IQ and IQ is what they try to measure. Additionally the evaluation of educational achievements include a self-reported part, which means that if Jews overestimate their math. talent more than other Americans (Americans of my age could not care less of mathematical talent), so they get higher educational achievement scores for the GWAS. Then one finds SNPs and this should lead to an error. You mention that the researchers of the field know this problem, I wrote that it would be amazing if they did not. The error must thus be intentional.

    Look again at the plots in
    http://rpubs.com/Jonatan/jewish_pgs
    There you have the HRS study, which was one of the cohorts. You notice that Christians and Jews do not differ so much in cognition, but they differ in PGS. This indicates that the PGS correlates with Ashkenazi ancestry. Look at Plot of PGS vs IQ in Dunkel et al study. See that Ashkenazi data points are more spread in the horizontal than in the vertical direction. That suggests that the PGS correlates with Ashkenazi ancestry.

    It is fine to discuss with you res, but do you and Merculinian quite understand what is asked from us? That is, James Thomson asked to break the paper into pieces. That means, opponenting it. Opponenting means making critical comments of the paper, not defending the paper and certainly not what Merculinian is doing, making critical comments of the opponent. Usually in opponenting the defender is the author. This is why I first assumed you must be the author as you defend the paper. But OK, it can be that some other person defends the paper, but then he should know very well real and correct answers to the questions, preferably he should not try to guess what the author did. If somebody other than the author takes the role of the defender, then he should be very polite to the opponents. That is a permanent rule. And he must not get offended. He is playing the role of the author, not playing himself. But maybe people who are not from academic circles do not know what opponenting is and what its goal is (it is for improvement of the paper). Or maybe James Thomson does not know that what he asked sounds to anybody from academic circles like he asked for opponenting Piffer’s paper. I had no time to look at Piffer’s paper as the easiest attack against it was the Ashkenazi Jewish PGS score and the PGS used in that score. But of course, I did this only as a favor to James Thomson, so not so much work from me can be expected.

    I will stop here, so do not bother to answer to this comment. There is one troll in this thread and I do not much like Unz trolls. This troll certainly does not understand the concept of opponenting and he is far too tiresome and not good enough for me to continue this exercise.

    • Replies: @j2
    , @res
    , @Merculinus
  245. j2 says:
    @j2

    Just a final comment.

    It is today so that American universities cannot be granted the usual assumption of honesty. They must be initially suspected of dishonesty and their work must be more carefully investigated.

    This is similar to what was with Soviet research. The fact that Soviets in the area of biology supported Lamarckism caused it that Soviet research of any field on any topic that had an association with anything important to the prevailing ideology had to be suspected of fraud.

    In the USA the situation is similar: American historians are unable to calculate correctly the death toll of Jews in the Holocaust, though it only requires basic addition and subtraction from accepted quotas. And American physicists cannot find anything strange in the collapse of the WTC towers. These two things destroy American credibility in exactly the same way as Lamarckism destroyed the credibility of the Soviet Union. More and more foreign researchers are skeptical of American results. This applies to any field and American researchers must accept it. Especially it applies to such ideological topics as the IQ of US Ashkenazi Jews.

    Whether American universities ever regain the trust that they one enjoyed will depend on the future. At the moment it does not look so bright. And this is not any more schizophrenic than it was to be suspicious of Soviet research, though the majority of it was good. Some was not.

    • Replies: @CanSpeccy
  246. CanSpeccy says: • Website
    @j2

    Whether American universities ever regain the trust that they oncee enjoyed will depend on the future.

    First they will have to give up on gender equity and racial equity in hiring and return to the criterion of merit, which is hardly conceivable. A system so corrupt as the Anglo university with its PC police, not only in the US but at the ancient English universities of Oxfraud and Scambridge, cannot be reformed without a revolution.

  247. res says:
    @j2

    Your Dunkel link is broken. Here is a working version.
    https://www.researchgate.net/publication/330601752_Polygenic_Scores_Mediate_the_Jewish_Phenotypic_Advantage_in_Educational_Attainment_and_Cognitive_Ability_Compared_With_Catholics_and_Lutherans

    You completely failed to understand my previous comment. The PGS used in the WLS was based on the Lee et al. 2018 GWAS of EA. Here is the documentation for the PGS used from the WLS site:
    https://www.ssc.wisc.edu/wlsresearch/documentation/GWAS/Lee_et_al_(2018)_PGS_WLS.pdf

    Whatever your credentials, your knowledge of this field is obviously limited. You would be better served by at least spending some time listening to people who know more and are patiently documenting their statements for you rather than just endlessly repeating your own (often wrong) talking points.

    And how can you honestly make a claim to be honestly opponenting a paper when you seem both not to understand the paper and unwilling to make an effort to do so. It seems that you just saw an opportunity to complain about a topic of interest to you (and you had to go to a barely relevant reference of Piffer’s to do that) and took it.

    Writing a 700 word comment then concluding with “so do not bother to answer to this comment” is more than a little rude. But I am pretty much done here given that you appear not to understand the points I am making and unwilling to make any effort to change that.

    P.S. I wasn’t aware of how Lee et al. make their PGS data available. That document provides a good window into the process.

    Due to IRB restrictions, it is not possible to release summary statistics for more than 10,000 single nucleotide polymorphisms (SNPs). Therefore, researchers with access to the individual-level genotype data cannot reproduce the polygenic scores from publicly available summary statistics (https://www.thessgac.org/data). As a partial remedy, we are releasing the polygenic scores directly to researchers (but due to the restrictions, we cannot release the underlying SNP-level weights themselves).

    Does anyone know what IRB restrictions prohibit the release of summary statistics? Or which restrictions prohibit the release of the underlying SNP-level weights?

    • Replies: @j2
  248. j2 says:
    @res

    You just copied this text:
    “We constructed polygenic scores for individuals of European ancestry in two prediction cohorts: the National Longitudinal Study of Adolescent to Adult Health (Add Health, n=4,775), a representative sample of American adolescents; and the Health and Retirement Study (HRS, n=8,609), a representative sample of Americans over the age of 50.”

    This exactly expresses that there are American Jews in the sample. Otherwise it would not be representative, and we know that the HRS study had 153 Ashkenazi Jews. As the expectation must have been that Ashkenazi Jews have a higher IQ average (or do you think the researchers had never heard that claim?) therefore they measured a mixed population of two populations with clearly different IQs in a GWAS that produces a PGS for a proxy of IQ. That must introduce an error. It is a wrong practice. As you state that they did know the problem, it is an intentional error. That is, there is a serious reason to expect fraud, though a small possibility that it is lack of competence.

    This is a fatal error in Dunkel et al and because Piffer used the same PGS it is a fatal error in Piffer.

    You have agreed that mixing different populations in the respect to the measured parameter is a known error and researchers (should) avoid it, but from the text you copied it follows that they did not.

    “Writing a 700 word comment then concluding with “so do not bother to answer to this comment” is more than a little rude. ”

    You are far too sensitive. Bye now, regards and good luck.

    • Replies: @res
  249. @res

    Res, I know these links.

    Perhaps we just have different perspectives on the Piffer paper. But it is okay for us to feel passionate about a topic.

  250. res says:
    @j2

    You really don’t understand. Pathetic.

    • Replies: @j2
  251. res says:

    For anyone who is interested in the genetic distance (Fst) of Ashkenazi Jews from European populations, here is the relevant excerpt from Dunkel:

    Ashkenazi Jews and non-Jewish Caucasians have been found exhibit relatively low levels of genetic differentiation. Tian et al. (2008) found that Ashkenazi Jews exhibited FST values ranging from .0040 when compared with Italians to .0144 when compared with Basque (across eight Caucasian populations, the unweighted FST average is .009). This means that Ashkenazi Jews exhibit little genetic differentiation, relative to non-Jewish Caucasians (FST values ranging from 0 to .05 correspond to little genetic differentiation; Hartl & Clark, 1989). Values this low also correspond to negligible amounts of prospective linkage decay because this parameter has been found to scale quite strongly with FST (Scutari, Mackay, & Balding, 2016).

    I was unable to find the Fst information in the Tian et al. (2008) reference given:
    https://journals.plos.org/plosgenetics/article?id=10.1371/journal.pgen.0040004

    But I did find it in Table 1 of this later paper by the same lead author:
    https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2730349/

  252. j2 says:
    @res

    “You really don’t understand. Pathetic.”

    I understand it very well. You are a poor loser and had nothing but insults as arguments in the end.

    By the way, the Fst distance between Finnish Saami and American Ashkenazi Jews is very likely to be somewhat higher than between Basque and American Ashkenazi Jews, in case the Fst distances of Ashkenazi to European population interest you and you want to have correct information. You find distances from Saami to other populations, for instance to Russians and then from Russians to Ashkenazi, it will to to about 0.02 or a bit more, like 0.023.

    • Replies: @res
  253. @j2

    I think I get what j2 is trying to say here. Sadly, I came to the conclusion that j2 is obsessed with the Jewish question and this is the only reason why we have to deal with his attacks.Obviously the thought that Jews are different or higher than other groups in some aspect is irritating him. This is why he is so fixated on Dunkel’s paper and considered Piffer’s study only with reference to his Ashkenaz PGS.
    Our guy’s knowledge of pop genetics is obviously very limited, and his reading (skimming) of Piffer’s paper and Lee’s GWAS is lazy and sloppy.
    This issue is called population stratification. J2 seems to think that because 153 out of 8,609 individuals are Jews, this creates a population stratification problem. That is, 1.8% of the population is Jewish. A very small percentage, unlikely to have a major impact on the GWAS results.
    But why is population stratification mud the waters? The rationale is this: 1) Two groups have different allele frequencies; 2) The two groups differ in the average GWAS trait (in this case, Education); 3) This difference is only due to the environment, or cultural transmission. Note that if 3 is false, population stratification ceases to be an issue, because the population differences identify causal effect of genes, and insofar as 3 is true, it muds the waters. Let’s assume the worst case scenario that Jewish have higher EA solely for cultural-environmental reasons.
    Now, I think j2 is exaggerating the problem because 1) 1.8% is unlikely to have a major impact on the GWAS; 2) The GWAS authors adopted strict population stratification controls, which would have picked up most of the genetic component related to Jews. Hence, the little confound that was left was reduced to a tiny, non-significant portion of the PGS. “We measure prediction
    accuracy by the ‘incremental R2’ statistic: the gain in the coefficient
    of determination (R2) when the score is added as a covariate to a
    regression of the phenotype on a set of baseline controls (sex, birth
    year, their interaction and 10 principal components of the genetic
    relatedness matrix).”
    Unfortunately for J2, Piffer computed different PGS. The PGS he used for the main analysis was not the EA_MTAG, but it was the PGS based on Educational Attainment only. So his contention that Jews score higher than others because the EA_MTAG includes self report measure of math ability and they overestimate this is false. Piffer found EA_MTAG and EA only PGS to be highly correlated (about 0.98 if I remember correctly), and he chose to use EA_MTAG for the charts displayed on this blog (supposedly because EA_MTAG captures more variance), but the PGS used throughout his paper is pure EA, which is not subject to the issue raised by j2, because it does not include a measure of self reported math ability.

    • Replies: @j2
    , @j2
  254. j2 says:
    @Merculinus

    “But why is population stratification mud the waters? The rationale is this: 1) Two groups have different allele frequencies; 2) The two groups differ in the average GWAS trait (in this case, Education); 3) This difference is only due to the environment, or cultural transmission. Note that if 3 is false, population stratification ceases to be an issue, because the population differences identify causal effect of genes, and insofar as 3 is true, it muds the waters. Let’s assume the worst case scenario that Jewish have higher EA solely for cultural-environmental reasons.”

    Your point 3) is not required. I will explain this still one time. This time try to understand it:
    You have property P what you want to measure.
    This property P partially depends on SNPs that lower or rise the property.
    Property P also depends on other things, so you do not get a perfect.

    You have two populations A1 and B which differ on property P and also on genetics, so there are SNPs that are not increasing or decreasing P but their frequencies differ on the populations A1 and B.

    You look for SNPs that correlate with P. Then you get SNPs that indeed to raise or lower P in each population, but you also get SNPs that do not raise or lower P in either population but are more common in one population.

    Thus, your set of SNPs that correlate with P coming out of this study include SNPs that do not correlate with P in either subpopulation. Next, take a population A2 that has the same value for P as A1, while B has lower. A2 will presumably give the same score for a PGS calculated from SNPs that correlate with P in each subpopulation A1 and B, but the PGS includes SNPs that only correlate with belonging to A1 rather than to B. As A2 is not quite A1, this score will be smaller for A2 than for A1. Thus, when you apply the PGS to thre three populations you get the order A1, A2, B. But if you swap A1 and A2 and make PGS that way, then the order is A2, A1, B.

    Now, this would be obvious to any researcher, but you do not seem to understand it.

    About your incorrect speculations of the type:
    “Sadly, I came to the conclusion that j2 is obsessed with the Jewish question and this is the only reason why we have to deal with his attacks.”

    keep your ideas to yourself. I explained my motivation and I did not have any other motivation than this: the article by James Thomson says, let’s break the paper into pieces (or something like this). I did not see the paper interesting enough for careful reading, but I concluded that from such a soft field like this, I can spot a fatal error in just a day or two. Most papers from soft fields have errors that in stricter fields would be considered hard. I know this very well because for 10 years I supervised researchers on a soft field. It is impossible to make precise research on soft fields, because the field just does not have a precise test of what is correct and what is not. For this reason there usually are such fatal errors. The people of the soft field do not care about them, just like you do not seem to care about this methodological error (you say, so what, only 1.8%, cannot have much effect. It is not if it has much effect, it is an error and in a hard field such an error is fatal).

    Thus, to find a fatal error from a soft field is not difficult, and it does not require reading carefully the whole paper, which I was unwilling to do as this paper did not seem interesting, it is similar to other Piffer’s papers that I have read carefully. But one day I could spare and spot one fatal error. Fixing even one fatal error in a paper improves it, so my opponenting is useful to the author. It is naturally the choice of the author if he fixes the error or not. Some are hardheaded and insist that there is no error, but they are warned: somebody else will probably find the same error and discard the paper as faulty. More clever authors do address the problem. They at minimum write an explanation why this is not an error. But when doing it, they should carefully consider if it is an error or not.

    In this case Dunkel et al paper has an error in this not exclusion of the Ashkenazi. The simple correction would be to recalculate the measure without Ashkenazi stating that they took the same criteria as in Lee et al (2018), but because they want to compare Ashkenazi to Christian, they have to exclude Ashkenazi before calculating the PGS. Lee et al (2018) would be good for other things, like for comparing White with Black, but not for comparing Ashkenazi with other White for this reason. So, this would be the correct correction.

    What you suggest is that the authors Dunkel et al would write: Ashkenazi are only 1.8%, so even though they should have been excluded, we just ignore this problem. It cannot make much difference. Then they would produce their difference between Ashkenazi and Christians. The problem with this comment it that the reader would say: hey, how much difference does it make and why cannot you just calculate the correct way in order not to have this source of error. Are you lazy or something? So, as you see, your correction would not work nicely.

    “Our guy’s knowledge of pop genetics is obviously very limited, and his reading (skimming) of Piffer’s paper and Lee’s GWAS is lazy and sloppy.”

    From the comments on other thread on Piffer it seems to me that most of the readers would like you and res just to disappear from dominating the discussion. All you can write are things like this. Then, considering that you still had not understood the problem and I had to explain it again, while I explained it maybe 5 times already, you are not too smart at all. And res made a checkmate of himself in the last discussion and then got a fury attack and started writing how much more he knows (I doubt it, at least he does not understand very fast). So, how about you two simply stopping commenting in the Unz. Nobody likes you making such comments, and you are not that clever at all. I did not reply to you to start with. It was you and res who started attacking me. I allowed you to have your discussion. Why did you start attacking my comments that were for James Thomson, not for two self-defined geniuses?

    • Replies: @res
  255. j2 says:
    @Merculinus

    The problem with you and res is that you two know so little of research that you did not even understand that what James Thomson asked were critical comments to the paper, in fact, to both papers Piffer and Dunkin et al. I found Dunkin et al more interesting to break into pieces. Giving critical comments is called opponenting and it is a practice very often used in research.

    The goal is then to give critical comments to the paper. It is not for somebody like you or res to defend the paper. It is also not for somebody like you or res to speculate the implications of the paper. It is to give critical comments to the paper.

    As Piffer’s calculations of the PGS that he uses are correct, they have been correct before, the criticism can only focus on the selection of the PGS. There were two changes in this PGS to the ones in his earlier papers that I had read and not been able to spot a clear error. One was that Finns had moved up. I discussed this many times. The other was Ashkenazi Jews, which was a new addition. The criticism should be from these if there were to be critical comments.

    Utu understood the task and made a very good critical comment.

    What you and res did was you attacked me, attacked an opponent. Why anybody with any understanding of research would do that? Clearly, you do not understand research. Probably you were not asked by the authors of James Thomson to defend the paper. As I understand, Thomson asked Piffer to write a paper to defend himself, and that is what one would expect. So, who the hell were you and what the hell did you two imagine you were doing?

    Very clearly, you two were two amateurs who think they contribute the best comments on the thread, they know the most and they own the thread and everybody should carefully listen what they say. This false belief caused you to attack me, mock me and you still have not stopped. Would you look at the mirror and think a bit what you were and are doing and why. You should not do it.

    • Replies: @res
    , @Merculinus
  256. res says:
    @j2

    As I said before, what matters is the distance from each of the populations in the study using the PGS to the British of European descent population in the study which created the PGS (Lee et al. 2018). Not the other between population distances.

    By the way, Finnish Saami are a well known outlier in European populations:
    https://en.wikipedia.org/wiki/Genetic_history_of_Europe
    Saying they are distant from Ashkenazi Jews says little about the distance between AJ and other Europeans.

    I am really curious how I have lost given that you basically have not even engaged with my points. Perhaps you are using “ad hominems per word” as a metric? You have definitely “won” by that metric.

    Or perhaps (probably closer to the truth) you are simply using “academic qualifications” as the metric of correctness.

    And this was a great example of projection. I am consistently amazed how well that simple idea explains so much human behavior. At least Freud got something right.

    You are a poor loser and had nothing but insults as arguments in the end.

  257. res says:
    @j2

    You have two populations A1 and B which differ on property P and also on genetics, so there are SNPs that are not increasing or decreasing P but their frequencies differ on the populations A1 and B.

    You look for SNPs that correlate with P. Then you get SNPs that indeed to raise or lower P in each population, but you also get SNPs that do not raise or lower P in either population but are more common in one population.

    You still fail to understand the simple point that the studies with different populations (e.g. WLS, HRS, Piffer) did not look for SNPs. Pathetic.

    Those studies used the SNPs found in Lee et al. 2018 which was confined to British of European descent.

    Once you start with a false premise, it does not matter how many words you write based on it.

    • Replies: @Merculinus
    , @j2
  258. res says:
    @j2

    Very clearly, you two were two amateurs who think they contribute the best comments on the thread, they know the most and they own the thread and everybody should carefully listen what they say. This false belief caused you to attack me, mock me and you still have not stopped. Would you look at the mirror and think a bit what you were and are doing and why. You should not do it.

    And a persecution complex as well. What a charmer. (and this after accusing me of being sensitive in comment 252, I guess that was just more projection)

    j2, how about you spend a few moments to respond to my substantive points. Your ad hominems are both repetitive and tiresome.

  259. @j2

    I am sorry j2 but your entire comments are OT. May I remind you that James Thompson invited readers to comment Piffer’s paper? Whereas you focused on Dunkel et al’s and dealt with Piffer’s results only as they related to Dunkel’s? You are in the wrong place, so go away and wait for a blog post about Dunkel et al’s paper to post your stuff.

    • Replies: @j2
  260. @res

    Yes, j2 is so ignorant of GWAS literature that he doesn’t know the difference between training and prediction samples. I checked Lee et al’s paper again and this confirmed that the WLS and HRS were used to predict polygenic scores that were constructed using different samples. That is, they simply validated previously constructed PGS. Hence the presence of Jewish in the prediction sample does not affect the PGS. It would have been a problem if the primary GWAS had included a substantial number of Jewish individuals, but the primary GWAS was “All association analyses were performed at the cohort level in samples restricted to European-descent individuals.”
    So again, j2 is wasting our time by constantly misunderstanding references and concepts…
    .

    • Replies: @j2
    , @res
  261. j2 says:
    @res

    You copied the text how Lee et al (2018) was formed from two American populations of European descent. One of these populations was the one in the HRS study. We know exactly that that study has 153 Jews. The other population was representative of White Americans, so it also had American Jews, as they are White. This means that the PGS of Lee et al (2018) did contain American Jews. And this is the PGS used by Dunkin et al and it is also the PGS used by Piffer, even if Piffer did not take the Edu3_MTAG but the other one. There were two PGS versions plotted in the HRS study, so there were two, both formed from the same samples.

    This inclusion of the Ashkenazi Jewish population introduces an error. Though the number of Jews in a representative While American sample is somewhere around 4% (3% of the whole population), it is not at all clear how big influence this has. It very possibly has a very minor influence to other populations than Jews, but it may very well have a large influence to the PGS score of Jews.

    Let us stop this finally. There is an error. It is not a typo or other minor error, so it is what is called a fatal error (serious error, major error) in hard sciences. That means that it should be corrected and this is totally regardless of how large the effect is. And we do not know how large the effect is.

    So, there is an error. I tried to find an error and found it, so I contributed positively to Dunkel et al and Piffer papers. You, on the other hand, have made zero contribution to these papers. You decided to have your private discussion on the thread, you did not even try to find any errors. I do not mind you having a private discussion, but you should not disturb people who try to find errors.

    That is, let’s break the paper, means lets find serious errors from the paper. This is done for giving feedback to the author for him to improve this paper and later papers. It is a favor to the author if some experienced researcher like me cares to look for errors from somebody’s paper. I know Piffer needs such help because he does not seem to get his papers accepted by the IQ community, so naturally he would like to sharpen his arguments and to do that he should ask for somebody to read his works and find errors. Every researcher would do just that.

    But you two. You just do not understand anything of this why it is done, what is the goal, how you should help. In order to help you should have looked for errors. Instead, what you do. You, res, pretend to be a small girl who is always offended by so called ad hominen attacks (which I have not made a single such attack, I do not do such). Your friend invents insults and has a really hard time understanding anything. I wrote to him that his logical thinking is not good enough for a researcher. It is not, if I have to explain the same simple thing six times, then it is not good enough. Neither is your logical thinking good enough for a researcher. You understand too slowly. I have to explain too many times and still you two do not get it. That cannot be in research.

    So, let you two stay away from research. Do what you want but do not disturb researchers. You are simply not on the level for that.

    • Replies: @res
  262. j2 says:
    @Merculinus

    Lee et al (2018) was done on an American sample of European descent. It was a representative sample and therefore it did contain Jews. Then Piffer used that PGS in order to calculate a score for Ashkenazi Jews. That is wrong, so he has an error.

    Lee et al (2018) can be used for calculating scores for Europeans, American Whites, Blacks, east Asians and so, but it cannot be used for comparing Ashkenazi Jews with other White ethnic groups. It is not made for that. How large the result of this error is, is irrelevant because errors should be fixed. I do not say it is a very large error, I simply helped Thomson in finding a serious error. But the error may be quite large in the score for Jews. It hardly changes the scores of other ethnic groups and this is why the correlation with the old PGS is small. This is a result of having only few Jews. But for these Jews the score may change considerably. This we do not know and as we do not know, we cannot assume it is small.

    Now, let us stop Merculinus. I really do not care to discuss with you. I have to explain the same things too many times. I prefer to discuss with polite people who understand things fast.

    • Replies: @Merculinus
  263. j2 says:
    @Merculinus

    There is a reason to suspect that the PGS may be incorrect. You cannot start by assuming that the procedure is correctly explained. The PGS of Lee et al (2018) is the PGS_EA3_MTAG which gives hits only tot eh Wisconsin study in google. There is some relation with this study and this PGS and if there is reason to expect something, then we cannot assume that the procedure for making this PGS is as told. It is American science on a dubious field. That alone means that it is dubious in the world scale.

    The reason to suspect that the PGS does react to Ashkenazi ancestry are the three plots of the HRS study where Christians and Jews differ much less in congnition than in the two PGS scores.

    Your conclusions of if there is an error or not, are without any special value. It is for James Thomson and the authors to see if there is an error, and if at least some text should be added to the papers to clarify the issue. Those are sufficient results from opponenting a paper. So, I have done what was asked. But you two? What in the world do you thing you are doing? Is your goal just to demonstrate that you have no idea of what opponenting means, or that how difficult it is for you to understand the problem in the paper? Or that how stubborn you can be and how many insults you can invent?

  264. res says:
    @j2

    Instead, what you do. You, res, pretend to be a small girl who is always offended by so called ad hominen attacks (which I have not made a single such attack, I do not do such).

    You are shameless. I gave a specific example of one in my comment 203.

    I exactly know that you are one-two generations younger than me and there has been all this brainwashing, so how could you possibly know what scientific truth means. You do your way, as dolls pulled by some doll master byu strings, but I guess it has to be so today. Maybe when you, res, grow a bit older, you will understand things you do not understand today. (Please, do not reply before you have reached my age and have some minimum level of wisdom. You are incredibly stupid now, only you do not see it, nor can anybody tell it to you, but I can see it, I cannot help you now, just trust me on this)

    So I guess we can add liar to your list of qualifications.

    • Replies: @j2
  265. j2 says:
    @res

    That is not any ad hominem attack. That is a description people younger than me, who can be quite intelligent in the normal sense, but who do not understand that they are being manipulated. One manipulation is this very high IQ of American Ashkenazi Jews. Read American Meritocracy article by Ron Unz and you see how good American Jews today are. The high IQ that once was, was almost centrainly a result of selection, just as Lynn concluded, but American scientific media and popular media spread such claims like that there is a heterogygote advantage in some Jewish diseases (it can be shown that there is no such advantage since the carrier ratio and prevalence ratio give just the normal relation). Because of this, one should look if there is any error in the definition of the PGS for Ashkenazi Jews.

    If you think that my text was an ad hominem attack against you, then you really are like a sensitive young girl. You sound younger than me and not like real researchers of any real field of science. I think we can conclude it is so. And if you are real researchers, why do you, res, claim you do not have a Ph.D.? That is the beginning stage to be a researcher, you have to have that to start with. You say you do not have, so you are not a researcher. Merculinus writes mostly insults and takes a too long time to understand a simple problem, so not a researcher of any real field.

    And I think you would have done very wisely if you had not answered to that wise comment from me, before you have credentials so that we can both understand what opponenting a paper means.

    I do not lie, I do not make ad hominem attacks, but you have some serious problem in understanding what people mean. You better work on that.

    • Replies: @res
  266. @j2

    I was once told that a village idiot keeps making the same mistake and calling others dumb because he is so dumb that he does not look at the mirror…worse, he does,but he is so dumb that he does not pass the mirror test.
    j2 has NOT read Lee et al.’s paper. He lacks the basic research skills, so instead of reading the supplementary info in Lee et al., he googled EA_MTAG and found “hits” to Wisconsin study, so his associative cortex made the association EA_MTAG-Wisconsin. Since his thinking does not go beyond the associative level, he thinks EA_MTAG PGS was based on the Wisconsin sample. Despite us explaining to him the difference between prediction and discovery samples, he keeps confusing the two. The Add Health and the HRS are not part of the discovery sample which was used to construct the PGS. Lee et al’s paper explicitly says that these two samples were excluded. They were used to validate the PGS. So now let’s see what the discovery samples consist of.
    If j2 had read the Lee et al’s supplementary file, suppl. table 16 (https://static-content.springer.com/esm/art%3A10.1038%2Fs41588-018-0147-3/MediaObjects/41588_2018_147_MOESM3_ESM.xlsx) he’d have realized that tha majority of the sample is not American. The majority of the sample is from Europe (UK and Estonia).* The “primarily EU” is the 23andMe. Supposedly this means 80% US (23andMe ships to lots of countries so this is my best guess, I couldn’t find a precise figure). So let’s say 290K from the US+5,690+7,789+14,562= about 320 K from the US, out of 932K. So only one third from the US. Assuming the percentage of Jews in the US is 1.4%, we have to divide this by 3 to come up with the % Jews in the total sample: 0.5% at most. Since the % of Jews in the UK and Estonia is much less, it’ll not significantly increase the percentage, or 0.1% max.
    0.5% has minimal impact on GWAS results. Add to this the controls by Lee et al. for population stratifications via principal components and you see this becomes non-significant.
    *
    Primarily US 365,538
    United States 5,690
    Estonia 36,631
    United Kingdom 6,065
    United Kingdom 8,535
    United States 14,562
    United Kingdom 15,941
    United Kingdom 19,193
    United Kingdom 442,183
    United Kingdom 8,094
    Scotland 1,841
    United States 7,789

  267. res says:
    @Merculinus

    I looked into this more thoroughly, and j2 is at least partially right (I think by accident, doubt he looked at the SM below). Lee et al. 2018 does include both the WLS and HRS in their training set (though I don’t believe that is apparent from the paper).

    The cohorts used in Lee et al. 2018 are described in its Supplementary Table 16 Cohorts and Supplementary Table 1.1 Cohorts of Okbay et al. 2016.

    Here are links to sources for the Supplementary Tables for both of those papers:
    https://www.nature.com/articles/nature17671#supplementary-information
    https://www.nature.com/articles/s41588-018-0147-3#Sec34

    There were 71 cohorts used in all (including WLS, HRS, a couple of Finnish samples, and much more).

    The UKBB accounted for almost half of the sample so that portion should be robust (and was what I was focusing on). But almost a third of the sample is from 23andMe (self reports!).

    One important thing to note about the Lee et al. 2018 methodology is this point made by Steve Hsu:
    https://infoproc.blogspot.com/2018/07/ssgac-ea3-genomic-prediction-of.html

    Although the study used over a million genotypes, the data had to be aggregated across many sub-cohorts using summary statistics only. This does not permit the L1-penalized optimization we used to build our height predictor.

    I am not sure all of this matters given that Lee et al. 2018 seem to be well accepted, but it is good to be correct. If j2 has an issue with the PGS or SNPs being used this is the paper he should be criticizing.

    P.S. If I were doing research in this area I would be tempted to use the Lello (Hsu) et al. 2018 EA PGS instead (Claimed 9% variance explained for EA, compared to 11-13% claimed by Lee)). The training population is the UKBB (only, I believe) and they include out of sample validation (at least for height, not sure about EA).

    Abstract
    We construct genomic predictors for heritable and extremely complex human quan-titative traits (height, heel bone density, and educational attainment) using modern methods in high dimensional statistics (i.e., machine learning). Replication tests show that these predictors capture, respectively, ∼40, 20, and 9 percent of total variance for the three traits. For example, predicted heights correlate ∼0.65 with actual height; actual heights of most individuals in validation samples are within a few cm of the prediction. The variance captured for height is comparable to the estimated SNP heritability from GCTA (GREML) analysis, and seems to be close to its asymptotic value (i.e., as sample size goes to infinity), suggesting that we have captured most of the heritability for the SNPs used. Thus, our results resolve the common SNP portion of the “missing heritability” problem – i.e., the gap between prediction R-squared and SNP heritability. The ∼20k activated SNPs in our height predictor reveal the genetic architecture of human height, at least for common SNPs. Our primary dataset is the UK Biobank cohort, comprised of almost 500k individual genotypes with multiple phenotypes. We also use other datasets and SNPs found in earlier GWAS for out-of-sample validation of our results.

    What we really need is more large sample studies like the UKBB with good phenotypic information.

  268. res says:
    @j2

    I exactly know that you are one-two generations younger than me and there has been all this brainwashing, so how could you possibly know what scientific truth means. You do your way, as dolls pulled by some doll master byu strings, but I guess it has to be so today. Maybe when you, res, grow a bit older, you will understand things you do not understand today. (Please, do not reply before you have reached my age and have some minimum level of wisdom. You are incredibly stupid now, only you do not see it, nor can anybody tell it to you, but I can see it, I cannot help you now, just trust me on this)

    Q.E.D.

    P.S. And for a bonus: “you really are like a sensitive young girl”

    • Replies: @Merculinus
  269. @res

    Yes I double checked, the Add Health and the WLS are present in the discovery cohorts. This doesn’t change my conclusion that the American sample is less than 50% of the total discovery sample, and since in the UK or the other countries from which 23andMe samples come the percentage of Jews is around 0.3%, the Jews in the sample including Americans will be less than 1%. This has a non-significant effect on the PGS, especially after accounting for population stratification (not sure if the PCs used by Lee et al. pick up of the Jewish ancestry though).
    Of course it would have been better from a methodological point of view to leave out Jews from the European-descent cohorts. I am not sure why Lee et al. did not do this, maybe politically correctness or simply sloppiness.
    23andMe 23andMe, Inc Genomics company Primarily US 365,538
    Add Health National Longitudinal Study of Adolescent to Adult Health Family-based United States 5,690
    EGCUT Estonian Genome Center, University of Tartu Population-based Estonia 36,631
    ELSA English Longitudinal Study of Ageing Population-based United Kingdom 6,065
    FENLAND Fenland Study Population-based United Kingdom 8,535
    Geisinger Geisinger Health System Population-based United States 14,562
    GSII Generation Scotland: Scottish Family Health Study Population-based United Kingdom 15,941
    NORFOLK EPIC-Norfolk Prospective Population Study Population-based United Kingdom 19,193
    UKB UK Biobank (Full Release) Population-based United Kingdom 442,183
    UKHLS The UK Household Longitudinal Study Family-based United Kingdom 8,094
    VIKING Viking Health Study – Shetland Population-based Scotland 1,841
    WLS Wisconsin Longitudinal Study Family-based United States 7,789

    • Replies: @res
  270. res says:
    @Merculinus

    Our comments looking at the Supplementary Materials crossed.

    Just wanted to double check that you noticed the Supp Table 16 note:

    Our final meta-analysis was performed in 71 cohorts, 12 of which contributed new data for the present study. The remaining 59 cohorts (N = 199,819) were included in our previous meta-analysis (“EA2”). The 59 cohorts are a subset of the 64 discovery cohorts listed in Supplementary Table 1.1 in [17].

    Okbay’s Supp Table 1.1 shows quite the collection of studies across space and time with a variety of selection criteria. The HRS appears there.

    I agree with your points, but I wonder if it is better to focus on a single (or small number of) more uniform study (ies) rather than the motley collection we see in the meta analysis here.

    This is in addition to the loss of ability to use some techniques (e.g. compressed sensing) because they are working from summary statistics. The summary statistics based methodology sounds powerful, but by no means is a free lunch.

  271. Lee et al. 2018 are ambiguous. They list WLS and Add Health in table 16, and apparently HRS is part of EDU 2, but in the paper they say that they excluded HRS and Add Health from the discovery cohort.
    “We constructed polygenic scores for individuals of European ancestry in two prediction cohorts: the
    National Longitudinal Study of Adolescent to Adult Health (Add
    Health, n = 4,775), a representative sample of American adolescents;
    and the Health and Retirement Study (HRS, n = 8,609), a representative
    sample of Americans over the age of 50.
    Then:
    “All scores are based on the results from a meta-analysis that
    excluded the prediction cohorts.

    Regarding the Wisconsin cohort: “To obtain a better measure of prediction accuracy for
    cognitive performance, we used an additional validation cohort, the
    Wisconsin Longitudinal Study (WLS)”.

    So from the main text of the paper it appears that the WLS, Add health and HRS were excluded from the discovery cohort that generated the PGS.
    I guess one would have to email the authors to be absolutely certain. Regardless, since Americans in the sample are less than 50% of the total sample and probably closer to one third, and percentage of Jews is much less in Europe (0.3% in the UK), the Jews in the total meta-analytic sample are probably less than 1% and surely no higher than 2%. This has virtually no impact on the PGS, surely not enough to explain a modest fraction of the Ashkenazi PGS-European gap.
    Even in the WLS study used by Dunkel et al. the percentage of Jews is only 1%.

    Is it “better to focus on a single (or small number of) more uniform study (ies) rather than the motley collection we see in the meta analysis here”?.
    Probably not, because these would be under-powered. UK Biobank’s measure of cognitive performance is crappy, maybe with some scoring tricks one could improve its accuracy.
    Can you find the link to Lello et al.2018 (Hsu) paper that you mentioned? I will have a look.

    • Replies: @James Thompson
    , @res
  272. @Merculinus

    Check with the authors. I assumed sample of discovery and sample of testing were separate.

    • Replies: @j2
    , @j2
  273. res says:
    @Merculinus

    Can you find the link to Lello et al.2018 (Hsu) paper that you mentioned? I will have a look.

    Here is the preprint with full text: https://www.biorxiv.org/content/10.1101/190124v3
    Published version: https://www.genetics.org/content/210/2/477
    Some discussion in Dr. Thompson’s blog: http://www.unz.com/jthompson/heritability-lost-and-found/

    • Replies: @James Thompson
  274. @res

    Yes, I remember it now.

  275. j2 says:
    @James Thompson

    “Check with the authors. I assumed sample of discovery and sample of testing were separate.”

    That does not help.
    If there were Ashkenazi Jews in the sample that sets the weights for SNPs, then the paper needs a calculation that clearly explains why the error it causes should be negligible, if it is. That we cannot know.

    Consider a theoretical example, you have two SNPs that both raise educational achievements in every population and are identical in the effect, but the frequencies of the SNPs are 99% in the population A and 1% in the population B for SNP1 and 1% in A and 1% in B for SNP2. Assume A has average IQ 110 and B has 100, and the size of A is 1% and the size of B ia 99%. Then we have frequence*effect:
    0.01*0.99*110+0.99*0.01*100=2*(0.99*0.01*100) for SNP1
    0.01*0.01*110+0.99*0.01*100=1*(0.99*0.01*100) for SNP2
    So, SNP1 looks twice as good in the study, while it actually is equally good. This is because the IQ 110 is not caused by these SNPs in the PGS but most of the 110 is caused by something else.

  276. j2 says:
    @James Thompson

    James Thomson,

    The problem in Piffer’s method seems to be the following:

    The result of summing over the subpopulation (getting the country averages) should be exactly the same as if you create the predictor in a way where you assign the country average to each individual and in this way ignore their individual results.

    Therefore, what is being created is a predictor that from taking a person and looking at his genes predicts to which subpopulation he belongs and outputs as the predictor of his educational achievements the country average of educational achievements.

    The plot of the used educational achievement scores versus the country average PGS score should be a straight line. That is not any research result. It is a direct consequence of this way of forming the predictor.

    Plotting national IQs versus the country average PGS will not give an exactly straight line only because the national IQ and the educational achievement scores used in making the PGS are not exactly linearly related.

    This same problem is in Dunkel et al.

    Hope this helps you, more time I cannot spend on this issue, but, please, read this comment with thinking.

    • Replies: @Merculinus
  277. @j2

    You are the most ignorant person I’ve had to deal with. You didn’t read the paper where the author controls for genome-wide genetic distances between the 26 populations when computing the correlation between IQ and PGS? If what you say was right, the correlation between PGS and IQ should disappear after taking into account genome-wide genetic distances (Fst). But it’s unaffected.

    • Replies: @j2
  278. j2 says:
    @Merculinus

    “You are the most ignorant person I’ve had to deal with.”

    This is typical you. You only have these arguments and then you invent something that is clearly nonsense and try to give it as the answer. Earlier you invented that if SNPs are discovered without Ashkenazi Jews, then it would in some way matter if the prediction sample has Ashkenazi Jews. Of course, it would matter. What matters is especially the prediction sample. You just write the first thing that comes to your mind and then you add the first insult that you think of. This is all you can do.

    “You didn’t read the paper where the author controls for genome-wide genetic distances between the 26 populations when computing the correlation between IQ and PGS?”

    You do not understand what you read, that is the problem.

    “If what you say was right, the correlation between PGS and IQ should disappear after taking into account genome-wide genetic distances (Fst). But it’s unaffected.”

    No. It would not disappear if you taken into account genome-wide distances (Fst). This is a false idea.

  279. The prediction sample had only 1% Ashkenazi Jews and you’re ridiculous in arguing it has a significant effect on the results.
    What you’re arguing is that if Piffer had just plotted random SNPs instead of GWAS SNPs he would have gotten the same result. This is wrong, because controlling for random SNPs (that is, the Fst distances), does not diminish the correlation.

    • Replies: @j2
  280. j2 says:
    @Merculinus

    “The prediction sample had only 1% Ashkenazi Jews and you’re ridiculous in arguing it has a significant effect on the results.”

    It makes very little difference to the PGS score of non-Jews, but it makes all the difference to the PGS score of Ashkenazi Jews.

    “What you’re arguing is that if Piffer had just plotted random SNPs instead of GWAS SNPs he would have gotten the same result. This is wrong, because controlling for random SNPs (that is, the Fst distances), does not diminish the correlation.”

    No. You again do not understand. The SNPs are not random. They have been selected in the discovery stage to have influence to the measured parameter, educational achievement. Try to understand this:

    Assume that each subpopulation has very little variation in educational achievement. They all have exactly the country average score. This can happen in our little thought experiment. Now, make the prediction stage and see what you get. The populations have a bit different genes and differenct average educational achievement scores (a proxy for IQ). You do get a predictor this way. Your predictor takes the SNPs that the person has, compares it with the SNP frequencies of different populations, and predicts to you an educational achievement score that is the average of the subpopulation you mostly resemble (or a linear combination of several if you do not clearly belong to any subpopulation). Thus, for instance, let’s taken an Ashkenazi Jew, who because of long isolation have rather uncommon SNP frequencies. The predictor notices: this person looks like an Ashkenazi Jew, so hist IQ is 110. So, it gives you an educational achievement score of a Jew.

    Next, put back the subpopulation variance. If you now use the predictor for a particular person, this predictor can detect, you are not only a Jew but an exceptionally stupid Jew, so it gives you the score 90. This predictor then works as intended and very well.

    Next, play Piffer and take an average of all people in each subpopulation. By taking this average you exactly remove the step where you added the subpopulation variation. You are back in the predictor that gives to each person (or actually, each country) the average educational achievement.

    So, this is not any random SNPs. The SNPs are quite correct and good, and the predictor is quite correct and good, but as several subpopulations with different IQ and different SNP frequencies have been used together as the prediction sample, when you sum the country PGS scores, you actually have only a predictor that does exactly what you programmed it to do: you though it by the country average IQ scores, and it gives you country average IQ scores. That is, the country average PGS score you obtain is exactly a linear function of the average country educational achievement scores you used.

    Merculinus. You are not any mathematical talent. So, try to understand this simple explanation. Read it ten times until you understand it. It is correct.

    • Replies: @res
  281. res says:
    @j2

    So, this is not any random SNPs. The SNPs are quite correct and good, and the predictor is quite correct and good, but as several subpopulations with different IQ and different SNP frequencies have been used together as the prediction sample, when you sum the country PGS scores, you actually have only a predictor that does exactly what you programmed it to do: you though it by the country average IQ scores, and it gives you country average IQ scores. That is, the country average PGS score you obtain is exactly a linear function of the average country educational achievement scores you used.

    Your last sentence is true. The question is how is that linear fit accomplished if it is not making a prediction based on the actual genetics of EA/IQ? You seem to be arguing that can be done in a way unrelated to the genetics of IQ (see “chopstick gene” comments above) if those populations are represented in the PGS discovery populations.

    The problem is the country populations are much more diverse than the SNP study populations (e.g. Africans and Asians). So even if your idea was true (and various criticisms above related to the small size of the subpopulations and small Fsts involved make clear that is unlikely) it would not work for African and Asian countries at all since they are not represented in the SNP finding studies.

    The other alternative, that the PGS is representing something real about genetic differences in IQ, works (with some loss of accuracy) in those populations assuming there is sufficient similarity in the genetic architecture of intelligence between those groups–which appears to be the case.

    • Replies: @j2
  282. j2 says:
    @res

    As you, res, ask me in this thread, so I copy a text I wrote to my blog but have not put there. It explains my criticism. After this rather lengthy part I will address your question.

    Assume that scientists have found a set I of SNPs (short pieces of DNA where a single nucleotide is different, they are caused by mutations) that are linked to some property P. In this case the property is educational achievement and it is used as a proxy for IQ, but for this calculation P can be anything. The size of the set I is denoted by S(I).
    Each of these SNPs has an effect in educational achievement, positive of negative. We will denote the effect of SNP number i by wi. The test sample is used to determine wi. So far let us assume we do not know it.
    We have a set J of test people that we will use for making a predictor for P, that is, fixing the values of wi. Each person j in this set either has an SNP number i or does not have it. Let us denote this situation using a number mji. Then mji=1 if the person j has the SNP number i, and mji=0 if he does not. The size of the set J is denoted by S(J).
    When we have assigned values to wi, the predicted score for the property P for the person j is
    Ȓj=∑i mji wi
    We know the real value Rj of property P because we make tests for the tests people. We would like to find such weights wi that the predicted score Ȓj is as close to the real value Rj as possible. Naturally, if the size S(I) of I and the size S(J) of J are equal, we can simply solve the weights from a matrix equation. We set Ȓj=Rj for each j and solve
    Rj=∑i mji wi
    which in matrix form is
    R=MW
    and the solution is
    W=M-1R
    but we want to use a larger test set, thus S(I)<S(J) and the equation is over-determined. The reason why J must be larger is that the SNPs do not in reality predict P so well. If we solve the weights for one set of test persons and then for another, the values for the weights are different. In fact, we cannot demand that Ȓj=Rj for each j. The most we can demand is that the error
    E= ∑jd(Ȓj,Rj)
    is minimized. Here d(A,B) is the distance between A and B. It may be the eigenvalue |A-B| or the square (A-B)2, or something else depending on our choice of the distance.
    Finding weights that minimize the distance is a normal optimization problem and can be solved without any difficulties. We can allow the weights to have any real number values, or we can restrict the weights in any way we want. For instance, we can demand that the weights can only be integers or signed binary numbers (+1, -1, 0).
    OK, so that is how you would do a GWAS. You can do it a bit differently, but basically it is like this: you make a predictor for some data. The predictor looks at your genes and predicts what value you should get for some property P.
    Assume now that the weights have been found. Let us go to the problem in Piffer's method. Assume that the set of test people is an union of several subpopulations Jk, k=1,…, N, and that these subpopulations differ in gene frequencies so that the frequencies of all SNPs in I are not the same in each subpopulation. Assume also that these subpopulations have different average values for the property P. We denote the average value of P for the subpopulation Jk by Rk,ave. The subpopulation PGS score for the subpopulation Jk is
    PGSk=S(Jk)-1 ∑jϵ Jk Ȓj\
    The average P score for the subpopulation Jk is
    Rk,ave=S(Jk)-1 ∑jϵ Jk Rj\
    Piffer plots Rk,ave as a function of PGSk and gets an almost straight line. But notice that
    PGSk- Rk,ave =(S(Jk)-1 ∑jϵ Jk Ȓj\) – Rk,ave
    and it can be expressed in two ways
    PGSk- Rk,ave =S(Jk)-1 ∑jϵ Jk ( Ȓj\ – Rk,ave\)
    PGSk- Rk,ave =S(Jk)-1 ∑jϵ Jk ( Ȓj\ – Rj)
    That is, the PGS score you get by averaging over Jk is exactly the same score that you would get if each person j has for property P the average value for P in his subpopulation. We can just as well think about Piffer's test in the following way. He has a test set of N subpopulations. There is zero variance for educational achievement in all subpopulations, but the subpopulations have different genes and different average educational achievements.
    Reflect on it for a while: what does the predictor use the SNPs for? It does not need them for the value it assigns to a person for the property P. That value is taken directly from the subpopulation P values. The SNPs are needed only for identifying the subpopulation where the person belongs to. The predictor could use SNPs that have no association with property P. The predictor looks at the SNPs in a person's genes. From these SNPs it can detect to which subpopulation the person most likely belongs to and it gives the country average value for P as the prediction for P to this person. If the person does not clearly belong to any subpopulation, the prediction is a combination.
    Think what happens e.g. in Dunkel et al (2019) test. They use PGS from Lee et al (2018), where the prediction test sample includes White Americans of European descent and a part of them (1% in the HRS sample) are Ashkenazi Jews. Dunkel makes just as Piffer, just as described above. US Ashkenazi Jews have an average IQ of 110 and US White Christians have an IQ of 100. The predictor can usually tell that if a person is a Jew from SNPs, as these populations are genetically different. To a Jew it gives the score 110, to a Christian it gives 100. This happens in any sample with White Americans. It does not need to be the test sample. The weights of the predictor are set so that when summed over a subpopulation it works as a predictor of the subpopulation and gives the subpopulation average as the prediction. Consequently, this predictor gives US Ashkenazi Jews a very high PGS score. It is built that way. This has nothing to do with the real influence of the SNPs to educational achievement. This result comes merely from the facts that the average IQs of these two subpopulations differ and that the genes of the subpopulations are different enough for the predictor to detect to what subpopulation a person belongs to.
    Did you get it? Piffer's, and Dunkel et al's, correlation is a mathematical consequence of the way the PGS is created. It is not any research result. It does not prove anything of the importance of genes to IQ. It is similar to a case where you put a stone in your pocket and later find it, or where you, at some less obvious step multiply your number by three and at the end show, look, it is divisible by three. Of course there is a correlation between the country PGS score and the country average IQ, because the country average IQ highly correlates with country average educational achievement, and you created a predictor which should give a linear correlation between country PGS and country educational achievements.
    Some people are impressed that Piffer's results repeat in each test. Of course they must repeat. The reason there is the correlation is a mathematical identity. It must always repeat, but it does not mean anything. You would get the correlation even if the SNPs had no influence to the property P.
    The problem clearly is in Dunkel et al (2019) as they compare two subpopulations, which have different genes and different IQs, and these two subpopulations were both in the prediction sample. This study must therefore be discarded. It does not matter that Jews were only one percent: the PGS score for them comes from this one percent. Piffer's earlier papers have an IQ gradient in Europe. It seems that the prediction sample had multiple subpopulations from Europe. This is wrong. The last paper of Piffer does not show this North-South IQ gradient in Europe, because the sample was made from Americans of European origins. As they all live in similar circumstances in the USA, there is no IQ gradient. This naturally shows that Piffer's earlier papers were incorrect because of the prediction sample. The new paper is also incorrect: Piffer includes Ashkenazi Jews, which belonged as a subsample to the prediction sample.
    Can this problem be fixed?
    Basically one would think that if we simply count the number of SNPs affecting IQ without any prediction stage, then we should have a correct method. But is it correct and can we do this way? IQ differences between countries are at least partially caused by environmental factors. It would be wrong to create a predictor, which tries to claim that the differences are only caused by genes. I think fixing this predictor to actually say something of IQ differences between countries is hopeless. There is lacking a scientific basis for it.

    Now to your question. I copy it here:
    "Your last sentence is true. The question is how is that linear fit accomplished if it is not making a prediction based on the actual genetics of EA/IQ? You seem to be arguing that can be done in a way unrelated to the genetics of IQ (see “chopstick gene” comments above) if those populations are represented in the PGS discovery populations."

    This is treated in the text above. It could be done with SNPs that have no relation to educational achievement, but I do not think it is. It is done by genes that have influence to educational achievement and have been discovered by the discovery stage. They are perfectly correct genes, yet the same problem appears. It is just the summation. It reduces the genes to work only as a detector of the subpopulation.

    "The problem is the country populations are much more diverse than the SNP study populations (e.g. Africans and Asians). So even if your idea was true (and various criticisms above related to the small size of the subpopulations and small Fsts involved make clear that is unlikely) it would not work for African and Asian countries at all since they are not represented in the SNP finding studies."

    Yes. Piffer also calculates PGS scores for populations that were not in the prediction sample. This is in principle correct, but then there is the Fst problem: are the SNPs correct. I have not looked at this part of the question as I only tried to find one serious error in my very brief look to this issue as there were no more interesting articles in Unz at that time. You two made is so difficult to discuss any problems in Piffer's work that it really did not encourage me to spend more time on it. If you are honestly interested, I can naturally think about this. It is another issue in any case. Some other reason for the correlation there.

    "The other alternative, that the PGS is representing something real about genetic differences in IQ, works (with some loss of accuracy) in those populations assuming there is sufficient similarity in the genetic architecture of intelligence between those groups–which appears to be the case."

    I briefly touch this issue in the long text I copied. I understand what you mean: it could be so that some population has more positive SNPs than another and is more intelligent for that reason. However, I think that as these so far found SNPs explain only a small part and we know that there are other explaining factors (environmental), I think this cannot be correctly done at this stage of the research.

    I thank you for a comment that is posed in a normal scientific discussion manner. BR j2.

    • Replies: @j2
    , @res
  283. j2 says:
    @j2

    The formulas got messed up, there are lower and upper indices and the backslash should not be, it is from the font. Hope you can understand them. They are very simple.

  284. res says:
    @j2

    Yes. Piffer also calculates PGS scores for populations that were not in the prediction sample. This is in principle correct, but then there is the Fst problem: are the SNPs correct. I have not looked at this part of the question as I only tried to find one serious error in my very brief look to this issue as there were no more interesting articles in Unz at that time. You two made is so difficult to discuss any problems in Piffer’s work that it really did not encourage me to spend more time on it. If you are honestly interested, I can naturally think about this. It is another issue in any case. Some other reason for the correlation there.

    If the SNPs aren’t correct then the PGS should not correlate with IQ for those populations.

    My point was that if your explanation is the reason for the studied groups IQs correlating then the outliers should not correlate at better than chance.

    So I consider the outlier group PGS scores correlating with IQ as evidence in favor of the hypothesis
    1. The PGS (and Piffer’s various other SNP measurements) are actually measuring something relevant to group average IQs.
    and against your hypothesis (please correct if this is not a fair paraphrase)
    2. The PGS (and Piffer’s various other SNP measurements) are detecting non-IQ population variation which somehow maps to IQ across those populations (an example of what I refer to as the “chopsticks gene”) and do not in fact matter for IQ at all.

    It is possible that 2. is correct in limited form (there may be a small number of SNPs like that), but the PGS/IQ correlation appearing so strongly over ALL of the groups (including the outlier populations) is strong evidence that 1. is the more important factor.

    Lee et al. 2018 used further methods to look for causal SNPs in particular as well. Recall that Piffer also replicated his work with those. IMHO the argument is even stronger in that case.

    Merculinus also makes some good points. It might be helpful if he could summarize them in a less contentious summary comment. That would actually help me too. Each of the three of us has made over 40 comments in this thread and it is hard to keep track of everything.

    • Replies: @j2
  285. j2 says:
    @res

    “If the SNPs aren’t correct then the PGS should not correlate with IQ for those populations.”

    I give a simple example: we have SNP1 and SNP2 which both affect IQ in the same way, but population 1 has only SNP1 and population 2 has only SNP2. Yet population 1 has a higher IQ than population 2 for environmental reasons. Then a predictor seeing that a person has SNP1 but no SNP2 in his genes, predicts correctly that he has the higher IQ, but it has nothing to do with the effect of these particular SNPs. Here you have PGS correlating with IQ, while the SNPs are not the reason for this IQ difference. You could just as well have SNP2 and SNP2 having very small or zero effect on IQ, while the (environmental) IQ difference can be e.g. 20 points.

    “My point was that if your explanation is the reason for the studied groups IQs correlating then the outliers should not correlate at better than chance.”

    It is a long time since I read Piffer’s earlier papers and first one should check if he has not combined several populations, including non-European ones e.g. to his first 7 SNP PGS and made a prediction stage, which would cause this problem and a correlation over all populations. That is, first one should check if there in reality were outliers in the first studies. If this is not the case, then one should look for some other reasons for the very good correlation.
    In the last study there were outliers as the PGS is from Lee et al (2018) and that PGS did not use as a prediction sample all populations that Piffer uses. I cannot say anything of this part at the moment. However, notice that in earlier papers Piffer has Italians below British, which shows that he had a North South gradient in Europe, but in the last he does not have it. This has an explanation that this difference was environmental: as Lee et al (2018) used an American sample of European origin people who in the USA had the same environment, the difference disappeared.

    “It is possible that 2. is correct in limited form (there may be a small number of SNPs like that), but the PGS/IQ correlation appearing so strongly over ALL of the groups (including the outlier populations) is strong evidence that 1. is the more important factor.”

    Let us hope so, but it should be checked. My limited goal was to find a serious error, that is, an error that needs to be fixed. It may or may not have effect. That is not important, as an error is an error and a paper is improved if an error is fixed. The mechanism I have described does cause an error (for any SNPs used, it does not need to be some bad SNPs, it is anything you use causes this error) if in the prediction sample are populations that differ in the measured parameter and in genes. This is the case for Dunkel et al and it is also the case for the single point of Ashkenazi Jews in Piffer. That is what I set to do: find one error, let the authors do what they like with it. It is possible that the outliers in Piffer do indeed show a real correlation, which is not caused by a mathematical property of a predictor. If so, there is a result. Yet, one should check if in Piffer’s earlier papers he did not make prediction over all subpopulations. If there is this result, then there is the question: we know that many environmental issues do influence average IQs. Why should a predictor which ignores these things, that we know should be there, give so good prediction? That
    question cannot be just wiped away. It must be explained before the result is convincing.

    “Merculinus also makes some good points. It might be helpful if he could summarize them in a less contentious summary comment.”

    I have developed some kind of allergy to Merculinus. I am sure you two can manage without me. So, all my best and goodbye.

    • Replies: @res
  286. res says:
    @j2

    In your initial statement, remember that I was talking about the PGS correlating with a population A which was not in the discovery study of populations 1 and 2.

    It is a long time since I read Piffer’s earlier papers and first one should check if he has not combined several populations, including non-European ones e.g. to his first 7 SNP PGS and made a prediction stage, which would cause this problem and a correlation over all populations.

    Remember, Piffer does not do SNP discovery studies. He uses the work of others (as he should, that is actually a good constraint preventing cherry picking). As far as I am aware he has only used studies based on European populations.

    If anyone knows of a good EA/IQ GWAS on non-Europeans I would be interested in hearing about it.

    The mechanism I have described does cause an error

    More accurately, may cause an error. That issue has been known about for a long time and the GWAS researchers make a significant effort to avoid it. A number of arguments have been given above (Fsts and small numbers of the more distant subpopulations, like Ashkenazi Jews and Finns) for why the problem should not affect this study. In addition I gave an excerpt from the methods portion of the Lee et al. 2018 Supplementary Material discussing how they tried to avoid population stratification in this comment from the other thread: http://www.unz.com/jthompson/piffer-kicks-against-the-pricks/#comment-3197992
    Also note the reference to a paper about the “chopstick gene” from 2000 there.

    • Replies: @j2
  287. j2 says:
    @res

    OK, res, I have to do something different now. For what my comments may be worth, you can look at the post
    http://www.pienisalaliittotutkimus.com/2019/05/08/piffer-fails-science-prevails/

    Just a post, do not be offended. It is for sure not ad hominem. Just a scientific observation.

    • Replies: @res
  288. res says:
    @j2

    Thanks for the followup. You say “It is for sure not ad hominem.” yet I see things like (emphasis mine):

    What is rather odd is that these supremacists accept Piffer’s plot that claims that Finns are more intelligent than other Europeans. Should they not object to that? After all, Mongol-hybrids, that is what the Nazis thought of us.

    If you pay any attention to my comments overall I think you will see I am no “supremacist.” Unlike you, I am quite comfortable with the possibility that both Jews and East Asians have higher average IQs than Europeans.

    You are reading far too much into Piffer’s work. I don’t think anyone here would claim this:

    But it is not justified to claim that Piffer’s results show that Finns have higher genetic IQ than other Europeans.

    I think we all realize the PGS scores are just estimates. And estimates have inaccuracy. Your statement is a strawman.

    Someday in the future I ask you to review the comments here and see which of the two of us was more obnoxious. Remember that you lit into me early because you were not distinguishing me from Merculinus. From your comment 224:

    About ad hominem attacks, they were started by Merculinius. When there is such a trolling person in the discussion it is difficult to keep it in a scientific level. Your comments were quite fine and I am sorry that I did not make a clear difference between you and Merculinius. He started with giving me a conspiracy theoretician prize. That is not a good start at all.

    You are simply wrong that I don’t understand the issue you are raising. I have talked about it in detail and even linked papers in the literature discussing it (the “chopsticks gene”).

    What you still fail to understand is my point about how a predictor suffering from that flaw and not truly predicting IQ at all would fail to correlate better than chance for populations not in the SNP discovery sample. I consider that decisive proof that your proposed flaw does not affect Piffer’s work in a material way.

    My background is engineering. One contrast between engineering and science or math is that in the former we are (IMHO) quicker to realize that one inconvenient truth trumps pages of mathematical manipulations.

    • Replies: @j2
    , @j2
  289. j2 says:
    @res

    I will answer to this comment as it seems to be about the subject matter.

    “What you still fail to understand is my point about how a predictor suffering from that flaw and not truly predicting IQ at all would fail to correlate better than chance for populations not in the SNP discovery sample. ”

    There is no reason if would only correlate by chance.

    You forget that this PGS is a valid predictor of educational achievement before you sum over a subpopulation. Therefore, for a person who has few IQ ioncreasing SNPs or many IQ decreasing SNPs the predictor will give low predicted educational achievement score.

    This means that for instance Sub Saharan populations, which are genetically far from Europeans will get low scores because their people will in average have only few SNPs that increase or decrease IQ. This largely explains why many non-European populations get lower scores than Europeans, and why the scores go down when we go further from similarity with Europeans. This is one basic feature of the predictor.

    The second basic feature is that the predictor predicts similar results to genetically similar populations, thus European populations not in the prediction sample would get similar results. The same is for any genetically similar populations. Thus, we should expect that African or Middle Eastern populations, which are genetically close, would also get rather similar results.

    Then there are East Asians. I think they actually have the same SNPs as Europeans and more of the good ones. This part is not an artifact.

    *******
    That was your main comment. Then you added a selection of claims and statements. I will comment some of them.

    I have no problem with Jews and East Asians having higher average IQs than Europeans. I find the question of differences in national average IQs is rather irrelevant. To have a so high IQ that it matters, it should be over 150, and to have a so low IQ that it matters, it should be below 60. But these country averages differ rather little.
    Yet, it can be a research question.
    – I estimate American Jews to have currently an average IQ of 103.5 because the most reliable tests point to that number. (The test giving 110 for Jewish and 106 average for White American Christians, as the European average is 99.5, setting the American Christian to that number gives 103.5 for Jews. There are several other indications, including 107.5 verbal IQ deduced by Lynn by his own study, it also gives about 103/104 for the total IQ. The figure 103.5 also agrees with Lynn’s estimate for Israel Ashkenazi Jews, as well as with his estimate for some Balkan Jews)
    – I also estimate that the American Jewish IQ was higher, around 110, some 20-30 years ago. I have written two-three posts of Ashkenazi Jewish IQ as it is a popular topic. This explains my interest in it. I have nothing against Jews or East Asians. I have hardly ever met either ones. However, there seems to be an effort to claim a too high figure for the American Jewish IQ.
    – I accept Lynn’s estimate 105 for East Asians.

    “You are simply wrong that I don’t understand the issue you are raising.”

    You write that having subpopulations with different IQs and different genes CAN cause an error. This CAN is incorrect. It will cause an error, in fact, this error does invalidate the result.
    If you understand the problem, then you understand that the subpopulation PGS score is exactly the same as what you would get if every individual in this subpopulation had the average country score of educational achievement, that is, there is no variation of IQ in this subpopulation. If this subpopulation was in the prediction sample, then the prediction error was minimized as much as was possible. For that reason we can assume that for subpopulations in the prediction sample, the prediction error is small. That means the following: given a sample person from this subpopulation, the prediction takes his SNPs and maps to them a predictor, which is close to the country average in educational achievement. It does not relate to the actual increases or decreases that SNPs cause (which are small in any case). It matches the person to a subpopulation and gives a predicted educational achievement score that is close to the subpopulation average score. This is an error and this happens necessarily. It is not that it CAN cause an error, or that we do not know if the error is large. The error must happen and the score given is largely incorrect.

    “Someday in the future I ask you to review the comments here and see which of the two of us was more obnoxious.”
    I already apologized to you once. You picked up my comment where I say that you are young. Maybe you found it offensive. So, fine. You may be a pensioner, so not a young person, and you may be a pensioned professor of IQ research, but in the beginning you did sound as a very young but too self-certain Unz commenter. You too much made the impression that you claim to know the field and everybody should listen to what you say. You also like to write things like, you are pathetic, and as you write in this way not only to me but to others, I doubt you should feel so hurt by a few words. I get a bit impatient when explaining the same thing for a long time to people who just do not get it. As I wrote earlier, my problem was Merculinus and you kind of joined in on his side. But I will try to avoid both of you in the future, so you will not have more problems with me.

    “If you pay any attention to my comments overall I think you will see I am no “supremacist.” ”

    I did not claim that especially you were. Unlike the supremacists, you did understand where the problem is, or at least you say you do. Thus, I could not possibly have meant you. But there are supremacists on Unz. Even in the other thread on Piffer there may well be supremacists. Some supremacists argued in a lengthy way how White people have invented everything in some other thread. In addition to these supremacists, there are some neo-Nazis and Hasbara trolls, and other trolls. Piffer’s result seem to especially appeal to supremacists and to hard-line HDB people.

    Very good that you are from engineering and not one of those damn softies. I have also an engineering background in addition to a science background, starting from making two MSc thesis and finishing as a professor in two technical (engineering) universities and one non-technical one. If you worked as a researcher in some more difficult engineering field, you know that it is true what I wrote of the opinion those people there have of soft fields. I personally do not share their opinions. I have worked with many soft field people and supervised quite many of them. But I never met somebody like Merculinus.

    Maybe you are nicer than Merculinus, but it seemed more like the old good cop bad cop trick with you two, so I finally decided not to make a distinction between you two.

    But now, let us finish this. I will not answer on this thread either. This IQ stuff is your field of interest. I am sorry I read the article by Thomson and wrote some comments, but it will not happen again.

    • Replies: @Merculinus
    , @res
  290. j2 says:
    @res

    “What you still fail to understand is my point about how a predictor suffering from that flaw and not truly predicting IQ at all would fail to correlate better than chance for populations not in the SNP discovery sample. ”

    I will still explain in other words the argument I had in the post about these outliers. You say you understand my arguments, but I am not sure, so I try to reformulate it.

    Consider Sub Saharan Africans. Piffer calculates a score using SNPs from Europeans. We would expect that the score is low for Africans as they are genetically far form Europeans.

    This score happens to agree with what the IQ of Africans could give. This is a coincidence.

    Why I say it is a coincidence is that it was possible that Africans would have developed their own IQ increasing mutations. If that had been the case, then Piffer’s low score would be wrong because these native African SNPs would have given Africans a high, even a very high.

    That was possible. The fact that it did not happen does not change the fact that is could have happened. There was nothing in Piffer’s method that could have given Africans a high PGS score in case it had been so that Africans have a very high IQ because of their own SNPs. This means that it was a lucky coincidence that Africans did not develop these SNPs. It was not evidence that Piffer’s correlation works. It was evidence that Piffer is a lucky guy. Or, he calculated PGS scores and noticed that they agree with reality and announced it as a correlation that explains the real IQ scores. It does not explain as the Africans are where they are only because Piffer got lucky.

    I hope this is now clear. It is written in the post, so you read it and I know you undersood it already, but just in case you did not, it is here again in a simpler formulation.

    • Replies: @Merculinus
  291. @j2

    Please help me j2, you’re so smart! I have just run an analysis but my results are too good, and would shatter all your beliefs about race and race denialists like you! There must be a mistake somewhere. I am sure you can spot it! Soft science people like me need the help of hard nosed guys like you!

  292. res says:
    @j2

    You forget that this PGS is a valid predictor of educational achievement before you sum over a subpopulation. Therefore, for a person who has few IQ ioncreasing SNPs or many IQ decreasing SNPs the predictor will give low predicted educational achievement score.

    This means that for instance Sub Saharan populations, which are genetically far from Europeans will get low scores because their people will in average have only few SNPs that increase or decrease IQ. This largely explains why many non-European populations get lower scores than Europeans, and why the scores go down when we go further from similarity with Europeans. This is one basic feature of the predictor.

    The second basic feature is that the predictor predicts similar results to genetically similar populations, thus European populations not in the prediction sample would get similar results. The same is for any genetically similar populations. Thus, we should expect that African or Middle Eastern populations, which are genetically close, would also get rather similar results.

    Then there are East Asians. I think they actually have the same SNPs as Europeans and more of the good ones. This part is not an artifact.

    So if I interpret this correctly, what you are saying is that despite the issue you raise the predictor still works across the different populations? It seems like you creating a very tenuous just so story to reconcile the predictor working across all of those populations without actually measuring what it is supposed to.

    By my engineering lights, if I have a predictor that works across all populations on Earth then that is indistinguishable in any practical sense from a predictor that is more theoretically satisfying. And in fact, that is the point where I start wondering about the correctness of the theory in question (as I think I have made clear already). “Proofs” of the inability of bumblebees to fly come to mind.

    P.S. You should not automatically make those assumptions about people in softer fields. In particular, I think people like Charles Spearman and Arthur Jensen make that quite clear. And I hope you don’t class the current generation of biologists doing things like GWAS and systems biology as being in a soft field.

    • Replies: @j2
    , @j2
    , @j2
  293. @j2

    Please, give up. You are uttering nonsense at every sentence, anyone with a minimal knowledge of population genetics can tell you’re not from the field and you’re badly in need of a remedial course in genetics.
    Now you’re turning to the argument that Piffer’s results were a matter of luck, which in this case means post-hoc.
    Piffer’s results cannot be post-hoc because they worked 3 times, in 2103, 2015 and 2019. They could have been post-hoc only the first time, but since the 2013 and 2015 paper had clear predictions, these were not post-hoc results.

    • Replies: @j2
  294. j2 says:
    @Merculinus

    “Please, give up. You are uttering nonsense at every sentence, anyone with a minimal knowledge of population genetics can tell you’re not from the field and you’re badly in need of a remedial course in genetics.”

    You are a troll. There are disinformation trolls in Unz, who pretend to be experts and try to control discussion. They usually come from one group of people.

  295. j2 says:
    @res

    “By my engineering lights, if I have a predictor that works across all populations on Earth then that is indistinguishable in any practical sense from a predictor that is more theoretically satisfying.”

    This is indeed a problem with engineers. They do not think of all possible cases. They are satisfied with something that seems to work. In engineering, the best is too good.

    It is so that a predictor works as it is intended to work. Just from the way a predictor is built follows that it should work quite well in prediction. But this does not mean it is an explanation of the phenomenon. It is a heuristic matched to the phenomenon. You remember heuristics from your engineering studies. A heuristic formula is an incorrect formula based on an incorrect theory but such that it works quite well in some situations. Physics has many such formulas. Engineers and practical physicists used to make them and use them. They worked quite well.

    The underlying theory that is the basis of the predictor is a linear theory of genes affecting IQ. If this theory is correct, then we should get rather correct not only average IQs but also SDs. It is difficult to say much of SDs of many developing countries as mostly SDs are not reported or the reporter standardizes the average and standard deviation for his purposes of making a comparison. This purpose is not to check the linear gene theory, so they do not give data for that. However, there is a claim that in some countries SD is smaller while the average IQ is larger. This situation could appear if some IQ increasing SNPs are nearly fixed. but there is too little data to say anything like this. Another reason for SD differences is the obvious: if parts of the population have different environments, you expect to see larger differences, if schools and everything else is similar, you expect smaller differences.

    Concerning Piffer, I say only the following:
    1) The data point of Ashkenazi Jews in Piffer’s last paper is incorrect as he makes the same error as Dunkel et al of comparing a sample of a subpopulation that was included in the prediction sample with another subpopulation that is genetically similar to a subpopulation in the prediction sample. This is an error and it means that Piffer fails, that is, the last paper should be corrected.
    2) The other part of Piffer’s claim is not rigidly refuted, but it is also not shown. You may find his argument convincing, it is your evaluation. Many people do not find Piffer’s correlation convincing. I do not. There are several problems with it: there are alternative reasons that are not ruled out.

    You think like an engineer: if Piffer’s correlation works, then the linear gene theory is correct. A mathematical argument showing that Piffer has an error must be false like the proof that bumblebees cannot fly. Only, there never was any such proof that bumblebees cannot fly. It is an urban legend. Mathematical problems are true problems and if it is simple mathematics, the problem is not imaginary and the argument is usually correct or easily shown incorrect. Piffer’s correlation may quite well work and still not give any strong reason to think that the underlying theory is correct. It is because there are too many unanswered questions of the theory.

    “You should not automatically make those assumptions about people in softer fields. ”

    Indeed, you do not understand what is written. You have this fixed idea that everybody is malicious and that all comments are criticizing you. It is like a victim complex: all are against you, while you are not doing anything wrong. So, people are not listening what you say, people are making ad hominem attacks against you, though you only want the best. I know this complex.

    If you read my text from the post, you notice that I refer to supremacist views. There are White supremacist, Jewish supremacist, hard field supremacist, Russian supremacist and whatever supremacist views. While these views are based on observations that are to some extent true (like that the sub-Saharan African IQ is around 70 at the moment, and that soft field people hardly ever can follow mathematical arguments of hard science, and I naturally do not mean my simple arguments that I try to write as simple as possible, everybody can follow them, if not, it must be intentional), it does not mean that the supremacist view is correct. I do not support any of these supremacist views. I have personally many positive experiences of soft field people. It is true thast they cannot follow some explanations, but then you just should avoid such explanations and explain differently, and if you have no common language, explain in the most simple way even though it may look like you underestimate the other side by explaining in so basic way, it is still the best way. I taught and supervised for ten years soft field people.

    In my opinion soft field people are quite intelligent in their ways, which is different from scientists, but in some things superior. But this can be also true of sub-Saharan Africans. You should not expect that they do well in IQ tests and school assignments, but they can be quite intelligent in a different way. Hard-line HBD people do not think so and in my opinion many of them have clearly supremacist views.

  296. j2 says:
    @res

    “So if I interpret this correctly, what you are saying is that despite the issue you raise the predictor still works across the different populations?”

    I still am not sure you understand the problem. You always talk about some chopstick gene. Try to understand this correctly:

    Assume you made a predictor, it can be linear or not. It has the property that given data of a person j (genes in this case) it outputs a prediction Pj. You know the individual score the person had for the measured property, Sj. Assume your person belongs to a subpopulation k.

    The difference PGS-average score for the subpopulation k is exactly the sum over the subpopulation of the prediction Pj minus the individual score Sj. And this sum is by a mathematical identity exactly the same as the sum of Pj minus the subpopulation average score.

    Because of this, you can replace your population you are testing with this PGS with another population where the individual genes are the same but each person j has the individual score as the subpopulation average. This gives exactly the same result.

    Next, thing what can your predictor be doing in this situation. If it works, it gives the average score of the subpopulation as about the same as what the actual average score should be. It will do so at least in the case you had a sample of this subpopulation in the prediction stage, as there you checked that your predictor works rather well. So, it does predict this subpopulation well.

    In order to give a subpopulation score correctly, in case the subpopulation has a higher score than the other subpopulations, your predictor must classify the test persons to the correct subpopulation. That is, your predictor uses the genes of person j only for assigning the person j to the correct subpopulation k. This happens with any genes. Not with some chopstick genes. It is exactly every single gene that has a different frequency in the subpopulation k than in the other populations. The predictor predicts to which subpopulation a person belongs to by looking at the set of SNPs he has and making a statistical prediction to what subpopulation he belongs. It cannot do anything else, it has no data for anything else.

    Now, to your question. What if the subpopulation was not in the prediction sample? What happens them? You can see the answer. The predictor tries to assign a person to a subpopulation that has genes as the person has. It does not need to be a real subpopulation, one of those in the prediction sample. The predictor does not keep any list of subpopulations that were in the prediction sample. It assigns the person to an imagined subpopulation that would have this kind of a score. The prediction sample populations set a direction to where the IQ increases and where it decreases and how much. The imagined subpopulation is on some point in this direction.

    Thus, assuming you had Finns or British with a larger number of positive SNPs in the prediction sample and some South European population with a smaller number of these SNPs, and the average country IQ scores are higher for the first two, the direction is that the score decreases if the number of positive SNPs decreases. That is, the predictor works rather well also on populations that were not in the prediction sample. It is not any random correlation. Yet, this is not a correlation that directly reflects the effect on IQ of these SNPs. When summed over a subpopulation, this predictor always uses SNPs only for identifying a subpopulation and the average score for this (imagined population for the predictor) subpopulation derives its value from the average scores of populations that were in the prediction sample.

    In order to make a real predictor that does not have the prediction step, we would first measure separately in each population to see what is the real effect of each SNP and then sum this effect for each person. As the total effect of these SNPs to the IQ is only some 4-20% or less, this real and fully justified predictor that was not created by using a prediction stage would predict incorrectly. It would predict that East Asians are only <1 point above Europeans, since the total effect of the SNPs to the intelligence is only less than one point. We could make this kind of a predictor to work, but first one would have to find all genetic sources of IQ and then this predictor could correctly give the genetic IQ differences, which still would not plot a straight line with measured IQ differences, since measured IQ differences depend also on the environment.

    I hope you now see the problem in Piffer's approach. If not, then I cannot explain it.

    • Replies: @res
  297. j2 says:
    @res

    I have this old teacher’s wish to get all to understand the issue correctly.

    A PGS is a predictor of educational achievement. It is a good predictor and it predicts well. Summing over subpopulations it predicts subpopulation average educational achievement well. If a predictos is well made, it works for all populations quite well. Thus, you get a plot of PGS versus educational achievement and this plot is very close to a straight line. It is not random correlation for populations that were not used in the prediction stage. This is how a predictor of any property should work.

    But a predictor is not an explanation of the property. It is a predictor and gets the values from (unexplained) individual scores for the property. These individual scores can have origin in education, genes, what ever. A predictor only matches these scores to genetic data that allows it to assign a good prediction to an individual. If subpopulations have different average values for the measured property, a correctly made predictor has this information in its weights. Therefore it uses for each individual the data from genes to predict inter-population and intra-population placement of the person with respect to this property. If we sum over intra-population variance, there is left only inter-population variance.

    So, the predictor works for ALL populations, but it does not in any sense imply that the gene data used by the predictor has any direct relevance to the measured property. In the case of PGS there is some small relevance: the SNPs do have a small effect on IQ, but when summed over a population what remains is the population differences in IQ. These population differences can have any cause.

    Read this many times before commenting anything. Try to understand it finally. There is no contradiction in what I wrote, nor is it in any way complicated. It is a simple thing about predictors. Predictors are used in many places in technical fields, and they have their properties. One must not confuse prediction with an explanation.

    • Replies: @j2
    , @Merculinus
    , @res
  298. j2 says:
    @j2

    There is a correlation between skin color and educational achievements. Thus, making a predictor from SNPs determining skin color will give a rather good Piffer correlation, which works for ALL populations and is not at all random. From that it does not follow that skin color genes influence IQ.

    • Replies: @res
  299. @j2

    Btw you’re also using SNP in the wrong way. You should replace SNP with ” allele”. If a population doesn’t have SNP, it means they’re monomorphic at that genetic site, hence they have only the ancestral or “wild type” allele, and not the derived (“mutant” allele. A SNP is a nucleotide substitution which may or may not be present in all human populations. Since we’re talking about SNPs shared by all the populations, at least within the GWAS sample, it’s wrong to say some populations have the SNP and some don’t. You should say, that some populations have more of the positive or negative effect alleles at some SNPs.
    Seriously, I mean it: I cannot discuss with someone who pretends to be an expert and lacks the basic training in my field and who would flunk genetics 101.

    • Replies: @res
  300. res says:
    @j2

    Do you understand what the “chopstick gene” idea is? See the paper I linked in one of these Piffer threads. It is one way the experts in this field refer to the problem of finding genes (SNPs) which are related to different populations rather than the trait in question. In other words, exactly the problem you have been going on about.

    For those who don’t know (and apparently, like j2, don’t want to follow links to find out), the “chopstick gene” refers to the idea that if you do a GWAS for the trait of “chopstick use” across worldwide populations you will get hits for “Asian” (i.e. cultural) rather than anything genetic which truly indicates chopstick use.

    I keep using the term because it is a concise way of referring to the population stratification problem j2 keeps raising, but can’t seem to be bothered to understand how researchers address.

    As for who is correct, I think we are at the “agree to disagree and wait to see who turns out to be right” phase. Because most of the things we discuss here do have answers which will be known–and with the speed of progress in genetics it might not take that long.

    Will Piffer’s bumblebee fly, or will j2’s mathematical manipulations keep it grounded? Tune in later to see.

    P.S. Some epic projection in j2’s comments. Do people not realize how informative such comments are about themselves?

    P.P.S. The “I am done now” grandstanding is fun to watch. Pro tip: you (and others here) should not engage in that unless you have more willpower than you (or I ; ) do.

  301. res says:
    @Merculinus

    Thanks for clarifying that. Not sure if I have done that anywhere. I know the difference between gene/allele/SNP, but occasionally am sloppy. For example, the “chopstick gene” I keep referring to, because that is how others refer to it colloquially. In reality what would be found is a variety of SNP variants (alleles).

    Seriously, I mean it: I cannot discuss with someone who pretends to be an expert and lacks the basic training in my field and who would flunk genetics 101.

    Understood. I’m a little surprised that j2 (being a retired academic) does not seem to realize how well the crankiness of your responses maps to that.

    IMHO it is a very different thing to be cranky when one knows an area and is right and to be cranky when one does not know the area and is wrong. Things become complicated when knowledge/correctness don’t map like that ; )

    P.S. I am using “cranky” in the sense of irritable or hostile. Not eccentric.

    P.P.S. For anyone who wants the definitions of genetics related terms, this is a good glossary:
    https://www.snpedia.com/index.php/Glossary

  302. res says:
    @j2

    There is a correlation between skin color and educational achievements. Thus, making a predictor from SNPs determining skin color will give a rather good Piffer correlation, which works for ALL populations and is not at all random. From that it does not follow that skin color genes influence IQ.

    That is actually an EXCELLENT example. And why controlling for population stratification is so important. But since the GWAS DO control for population stratification that should not be much of a problem.

    But assuming for a moment these controls weren’t done, what might we see. Here is a look at:
    Association of race and color with mean IQ across nations.
    https://www.ncbi.nlm.nih.gov/pubmed/17037466

    Abstract
    This study investigated the correlation of both race and skin color in the distribution of mean IQ for 129 countries with primarily indigenous populations. Skin color correlated most highly with mean IQ across the Caucasoid countries (r = -.86), somewhat less across the Mongoloid countries (r = -.66), and nonsignificantly across the Negroid countries (r = .06). When the Negroid and Caucasoid countries were combined, both race and skin color yielded high correlations with IQ (r = .87, -.95, respectively). When the Negroid and Mongoloid countries were combined, both race and skin color yielded high correlations with IQ (r = .91, -.91, respectively). When Caucasoid and Mongoloid countries were combined, skin color yielded a high correlation, but race did not correlate significantly with IQ. The greater importance of Negroid race was regarded as congruent with the 1998 generalization of Jensen that the genetic distance between Mongoloids and Caucasoid is less than the genetic distance of these two groups from Negroid.

    So if we used a skin color PGS rather as proxy for an IQ PGS we might expect strong correlations across all populations and most subpopulations, but not across Negroid populations. Which is a little like Piffer’s Figure 2 divided into three groups along the x-axis, except there the middle group also has a small correlation. (but see additional comments below)

    FWIW, here is a graphic giving an idea of the skin color IQ correlation:

    Notice that Piffer’s work shows three Asian populations with highest PGS (Figure 2) which is not what one would expect from skin color (reordered to correspond to IQ).

    P.S. One way of checking that the EA/IQ GWAS don’t have that problem is to notice that despite the strong correlations observed above, as far as I know, none of the EA/IQ GWAS have produced SNP hits which correspond to skin color (the genetics of which are fairly well known). Since skin color has a fairly small number of SNPs with large effect they should appear quite strongly in any EA/IQ GWAS subject to the issue j2 raises.

    P.P.S. j2, you don’t seem to realize that you are mostly arguing against the GWAS research rather than Piffer at this point. That science is rather well established now. Perhaps you could take your criticism to the dozens of coauthors of Lee et al. (2018) and see how they respond?

  303. res says:
    @j2

    FWIW, my background is in modeling and simulation.

    One must not confuse prediction with an explanation.

    Definitely a worthwhile point (and if Okechukwu happens to read this, that is what taking an opponent’s argument seriously looks like). But given such a good predictor one should seek to understand WHY it is so good. (not, the FUD of “it might be wrong”, but what mechanism–like skin color–allows it to give good predictions despite being wrong)

    Occam’s razor is useful.

    You raise an issue with population stratification causing non-EA/IQ SNPs to be found by the GWAS. But given that this is a well understood problem which GWAS are specifically designed to deal with through the use of relatively uniform study populations, controlling for principal components of population genetic variation, and post-hoc checks (see Lee et al. use of QQ plots in their Supp Materials which I linked elsewhere); I think it is reasonable to assume the results are correct (with respect to this issue) absent evidence this problem is (not just might be) present.

    As I have said multiple times before, that the predictor works well out of sample (i.e. in populations not included in the GWAS) is strong evidence it is capturing the true effect we are looking for.

    Again, you are arguing that the GWAS is incorrectly finding non-EA/IQ related SNPs. This is a valid concern. Which you should take up with the authors of Lee et al. (2018) and the authors of all of the other GWAS Piffer has used to demonstrate his idea over the years.

    Who knows, you might have found something which destroys the validity of all GWAS everywhere and no one else has ever thought of. That would make you famous. You should go for it.

    • Replies: @j2
  304. j2 says:
    @res

    I basically tried to find one serious error in Piffer’s paper. That error is the Ashkenazi Jewish data point and it is really incorrect since the PGS of Lee et al (2018) used by Piffer and Dunkel et al was created by a prediction sample that included two populations that differed by genes and by IQ. This causes an error (not: can cause an error, but causes it). With Dunkel et al we have to assume that et al must have been aware of the problem, so they did it intentionally. With Piffer no intention can be shown and I doubt there is any. It is simply an error. This is where my interest to the paper ended.

    The second question was raised by you and I offered some thoughts: why is Piffer’s correlation over many populations so good if there is no underlying real reason for it. I said I can think about it if you honestly are interested in it. So, to help you I gave it some thought and some comments about this: you may get good results with a good predictor, but prediction does not always imply an explanation. I really have to do something else and will not continue thinking of this problem any longer.

    GWAS is mainly used in medicine for finding people who are prone to some illness. In that usage a predictor is what they need. The predictor does not need to be an explanation of why these people are vulnerable to the illness. Thus, none of my concerns can affect this usage of GWAS. Application of GWAS to the study of an effect of genes and of finding genes that cause some trait is a different application. There one has to be careful, especially when comparing scores between populations. In GWAS for educational achievement there is the intra-population variance, which does not have this problem, but of course, a society is also often stratified. There are for sure ways to use GWAS so that the problem does not appear. I do not think I need to contact Lee et al and I have no need to become famous on pension, nor to publish scientific papers on pension.

    The other way, that is, of making a model that explains IQ, is of course possible. Many known factors affect IQ and one can try to make a sum formula. Then it cannot only be a set of SNPs affecting IQ. There are environmental factors. In some societies, like ours, environmental factors decide certain percentage, but this percentage is dependent on the society. That is, it is easy to describe a society where genes practically have no effect on the IQ of an individual. For instance, res sets up a dictatorship and decides for each person what IQ he must have. All IQs are between 60 and 85 as res wants a country where nobody is intelligent enough to rebel. So, how does res do it? He measures the IQ of each newborn child and if it is below 85, he kills the child. Thus, all who live have genetic IQ at least 85. Then res picks up a number between 60 and 85 for the IQ of the child and orders his doctors to cut off enough of the child’s brain so that his IQ equals this number. So, here we have almost 100% of IQ determined by the environment.

    This example shows that the society has an effect. The effect of the environment can be smaller or larger. If the society excludes girls from education, then the IQ of girls is lower, and so on. I doubt one can find a simple formula that correctly predicts IQ in a way that it lists the effects of all factors.

    I suggest stopping this discussion. So, please, do not make any more questions to me, nor claims that should be corrected or answered. Merculinus is quite happy to discuss this topic, I am not. You should take my announcements that we should stop as a hint that let us stop.

    As a stopping reply that does not need to be answered, I give: I wish you all the best and now the final goodbye.

    • Replies: @res
  305. res says:
    @j2

    I suggest stopping this discussion.

    Why? Have you realized you aren’t making persuasive points? Though some are creative–like the idea of distinguishing medical GWAS from EA/IQ because prediction matters more for those, not whether the underlying model is correct. But really, for medical GWAS you do want to understand the mechanisms because that might be useful for treatment or drug design.

    Regarding the issues about Piffer’s use of the Dunkel paper, I think the bigger issue there is uncritically using their 110 number for Ashkenazi Jewish average IQ. That actually makes Piffer’s results look worse than if he had used one of the lower figures out there.

    If you want to stop the discussion all you have to do is stop responding to me. Assuming you are actually going to do that this time, I will wish you well. Hopefully someday we will find ourselves on the same side of an argument. Perhaps then you might see my comments differently.

    • Replies: @j2
  306. j2 says:
    @res

    “I will wish you well. Hopefully someday we will find ourselves on the same side of an argument. Perhaps then you might see my comments differently.”

    I too wish you well and will stop this discussion. A student, who when leaving a session with the supervisor, mumbles “Pythagoras theorem is wrong” will be called back for a short discussion of the proof of this ancient theorem, and so do misunderstood clarifications even to a farewell to be set right by a short comment.

    You see your comments as correct, but someday perhaps you will understand my comment and realize that you are a one field supremacist, of which a special case is a hard field supremacist. Almost all expert of any field are. For what is a supremacist but a territorial animal, who protects its territory so that outsiders do not come and steal their women and hunting grounds. It is for the territorial instinct that he requires the knowledge of his narrow field, as it would be at all needed in solving the problem. For that is the source of all his words and his feeling of superiority: the advantage of the home ground is the same as supremacy.

    I would rather have you say: you stranger, who have traveled so far and wide, seen the drums of Thule and the charms of Timbuktu, do you have a cure for this young cowboy who is sure bound to die, for our medicine is too weak?

    • Replies: @res
  307. res says:
    @j2

    Well, that is the fifth comment from j2 suggesting we stop this discussion or claiming he will (and counting?). In this thread alone. The following post has additional comments like that.

    What part of: “Your issue appears to be with the GWAS (not Piffer’s use of it) so take it up with those researchers and maybe ask yourself why GWAS are taken seriously (in fields beyond IQ research) at all” is so hard to understand?

    P.S. I will conclude with a point which I think illustrates that despite his protestations I really do understand the issue j2 raises. Consider doing a GWAS of lung cancer. It is important to be careful about accidentally detecting SNPs associated with smoking. Although one can argue those SNPs are still relevant, the distinction is important. So when we look at lung cancer GWAS papers we see things like:
    https://www.ncbi.nlm.nih.gov/pubmed/28615365

    In a second GWAS, a SNP within the CHRNA3 gene was strongly associated with smoking quantity and nicotine dependence (15). The same SNP was also strongly associated with lung cancer. The results suggest that the variant on chromosome 15q25 confers risk of lung cancer through its effect on tobacco addiction. In contrast, a third study showed weak evidence that the 15q25 locus influences smoking behavior and is mostly directly associated with lung cancer (16). However, it should be emphasized that the later GWAS was conducted in cases and controls matched on smoking status, thus limiting variation between the two groups and the power to detect any smoking association. Further analyses from the same study suggest that SNPs and smoking have independent effects on risk. Together, these three studies unequivocally support the 15q25 locus as harboring susceptibility variants for lung cancer or smoking behavior.

    • Replies: @j2
  308. @Merculinus

    Merculinus, this explanation does not seem convincing.

    Do not forget that statistical models have a margin of error. So the 84 and 88 are in the ballpark.

    My argument holds. The Piffer model appears to have failed on a fundamental level.

  309. j2 says:
    @res

    I do not think you understand the problem. I will try to explain is again. (When you leave your friend on a street you say “see you later”, you do not shout after him “but you were wrong in this point”so that he has to return and answer. That is not a polite way to depart.)
    I copy a short text I wrote, it explains what the actual problem is:

    The problem in Piffer’s results was that he summed a GWAS predictor of educational achievement over each subpopulation and by doing so he got a predictor that gives exactly the same result was what he would get if each individual in each subpopulation had exactly the subpopulation average score in educational achievement. So, he could replace his test population with this kind of population. In this new population the predictor cannot do anything else than to identify the subpopulation to which a person belongs by using the SNPs. The educational achievement score comes directly from the subpopulation average scores and they have no direct connection with the SNPs used.
    Indeed, a fairly good Piffer correlation can be made by using SNPs that determine skin color, and skin color genes have no effect on the IQ and the educational achievement score is supposed to reflect.
    The case that some SNPs may not be related to the property that is measured is a known problem with GWAS and researchers usually manage to remove this problem. For instance, in a GWAS of lung cancer, some SNPs may reflect a smoking habit and not cause lung cancer, while others are linked to lung cancer. The usual fix is to remove the former ones. So, this is known. It is also known that if the predictor sample includes genetically different populations, then the predictor cannot be used for comparing others samples populations between each others, an error done in Dunkel et al (2019). Neither of these two known problems is the fatal problem in Piffer.
    Piffer’s problem is related to these two, but it is different. All SNPs in the PGS used by Piffer are linked to educational achievement, so they cannot be removed from the PGS, unlike skin color genes, which could be removed. The problem is that their combined effect is rather small and they do not explain educational achievement scores. The scores the PGS gives are not derived from the real (small) effect of these SNPs. The scores reflect the measured values of educational achievement. Consider an example. If education is changed in a country, educational achievement scores are often changed. The population is not changed, so genes are not changed. We would expect that the PGS score for this country reflects the genes and is not changed. Thus, the data point of this country moves away from Piffer’s straight line. This is not what happens. The PGS score for this country changes, though the genes do not change. It is easy to see that if this country is in the prediction sample, the PGS score mimics the educational achievement score because that is where from it is calculated: it predicts the score. It is not a true value derived from genes. But this can happen even if the country is not in the prediction sample: if a genetically similar country with a similar education change, is in the prediction sample, this country, being genetically similar to the one used to create the predictor, moves the same way. You see the problem here. The data points stay quite well on a straight line, but they can move along the line because of environmental reasons.
    ( So it is here, the results e.g. for Finland can move up or down, though the population does not change. The PGS score is not what it looks like, it is not a measure of Finnish genetics. The score is much more a measure of the Finnish school. So Piffer’s correlation is misleading.)

    How to fix this?
    I suggest that Piffer explains that the country average PGS is only a predictor, not an explanation. It gives correctly the observed trend in educational achievements and as such is a good starting point, but the new contribution is an improvement of this predictor. This part is missing from the paper, but adding it makes a nice result. Piffer selects a set of environmental factors that are known to influence educational achievement, makes an additive correction term to the PGS predictor as a sum of environmental factors, and matches it to a large set of populations. Then he can show that his improved predictor is better than the original PGS. He can estimate how much the educational achievement score can be affected by those environmental factors, and he can even give recommendations. As a conclusion is notices that this amount of educational achievement can be influenced by fairly simple environmental factors, but it is not the total effect of environment, since the trend itself has environmental causes, in addition to genetic ones.
    The second result can be investigating the trend: Piffer makes a cultural model where advanced culture spreads from some centers, like Europe/USA and China/Japan, and also from India, but there the cast system limits the effect to upper castes. The further a population is genetically from these centers is likely to correlate with educational achievements. This gives a predictor that only uses environmental factors. Piffer compares a trend of this predictor with the trend of the PGS predictor and concludes some carefully formulated statements of the role of genes.
    A third result can be obtained by looking stratification within a country. A prime example is India with its case system, but the class societies of England and France will do, as will the USA with its WASPs and Jews. Piffer looks at the problem that SNPs basically only identify a subpopulation and tries to find cases when this identification of a subpopulation shows differences in educational achievement that are clearly not a result of genes.
    That’s three possible result suggestions for Piffer. You probably can easily think of some more.

    • Replies: @res
  310. res says:
    @j2

    When you leave your friend on a street you say “see you later”, you do not shout after him “but you were wrong in this point”so that he has to return and answer. That is not a polite way to depart.

    Agreed. You might review your own comments with that in mind. Especially since I have never said I am done, or leaving this conversation, etc.

    • Replies: @j2
  311. j2 says:
    @res

    “Agreed. You might review your own comments with that in mind.”

    Yes. I know and admit that my comment was exactly what I condemned, and I knew it when writing the last comment. It condemned both sides, me not less and I added it just to explain why I still write a comment to you. We just do not manage to stop this discussion in a civilized way, but tomorrow I will be on a trip and will not write comments for a long time.

    The problem with this discussion from the very beginning is that I knew that you did not understand the problem, while you believed and assured me that you did understand the problem and implied that I did not. This is what caused me to think and express that you are very slow in understanding the problem and this discussion is pointless, maybe in an offensive way. So, my apologies for this, but I was and am correct on this: if you had understood the problem, you would have reacted differently.

    Piffer’s straight correlation line is not a correlation where an objective measure from genes is plotted against an objective measure of intelligence. It is a line where a mimicker (predictor) of educational achievement plotted against IQ. All we can say is that the PGS predictor is well made (this time, by Lee et al) and that one predictor of educational achievement (PGS) correlates with another predictor of educational achievement (IQ). But the result is shown in a form that misleads one to think that PGS is genotype and IQ is phenotype and this line says something of heritability of intelligence. PGS is expressed by SNPs, but it is a sum of weights and these weights do not derive from the number of good genes in an individual but from the scores of educational achievement.

    But then, if you have not tried to stop this discussion, there is no harm that I write this comment.

    • Replies: @res
  312. res says:
    @j2

    It condemned both sides, me not less

    It is worth considering who did it first or has done it more frequently before attempting to make a (false) equivalence. See your comment 249 and my comment 251 for what I believe to be the beginning of the “I am done. Don’t respond.” interaction.

    but tomorrow I will be on a trip and will not write comments for a long time.

    We shall see.

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