The Unz Review - Mobile
A Collection of Interesting, Important, and Controversial Perspectives Largely Excluded from the American Mainstream Media
 BlogviewJames Thompson Archive
Piffer's Equation Further Updated
🔊 Listen RSS
Email This Page to Someone

 Remember My Information



=>

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

Piffer Lee 2018 continental pops

In June 2017 I declared open season on Davide Piffer, inviting criticisms of his findings:

The official response to Piffer is: “publish, and then we will give you our comments in reply.” This will take time, but it is the traditional way of doing things.

The unofficial response is to encourage more criticism right now, because if the finding is the result of a simple error, it should be exposed and corrected as soon as possible.

It is open season on Piffer’s methods. Recruit critics and get them conduct their peer reviews right now.

We are facing a French dilemma: Piffer has an approach to the genetics of racial differences in intelligence which seems to work in practice, but should not work in theory. His technique appears to run against the general trend of genetic research, in that he appears to be getting good results predicting group differences in intelligence on the basis of just 18 SNPs, while genetics researchers are getting only reasonable results in predicting individual intelligence on the basis of lots of SNPs.

For example, people skilled in these matters tell me that they did an out of sample prediction in an independent but European population, and they got 4.8% of the variance, using all SNPs. That is the upper limit of prediction in a non-European population using all SNPs. Pfiffer used just 18 SNPs in non-European populations and his correlation is huge, which does not make sense.

Piffer explained how he was able to achieve his results:

These SNPs that explain variance within populations are markers of polygenic selection. They do not have to explain a lot of variance between populations, or even within populations. The polygenic evolution model predicts that a few SNPs will have frequencies correlated to frequencies of countless other SNPs. I just need to know the few most important SNPs to gather a signal and infer to the distribution of the other unknown SNPs.

If selection pressure acted on these 9 SNPs by driving their frequencies up in population A compared to B, then it has also done the same to other SNPs. We don’t need to know what these other SNPs are because theory predicts that they will have similar distribution.

So, now that the massive James Lee study has been published, where does this leave Piffer’s polygenic evolution model prediction?

By the way, Piffer publishes in the modern sense of that word: he posts up his findings, together with all the code he used to generate his new results, thus allowing all and sundry to see inside the closet, and to check his figures for errors. You can peer review it and tear it apart here:

http://rpubs.com/Daxide/374949

You can see the results for the 52 populations below:

Piffer Lee results for 52 pops

You can see the results for the major continental groups below:

Piffer Lee 2018 continental pops

You can see Piffer’s conclusions and cautions below:

Piffer conclusions on Lee 2018

In sum, Piffer has provided a further test of his approach. He cautions that some of the sample sizes are far too small. With any luck this can be dealt with by sampling more widely and in greater numbers. Larger samples may become available with time.

The general pattern is interesting, in that it is broadly in line with the expectations from intelligence testing drawn from country averages, and racial group averages.

Once again, in the spirit of the fearless examination of the intellect, I ask you to subject his work to merciless enquiry and savage criticism. Over to you.

 
• Category: Science 
Hide 101 CommentsLeave a Comment
Commenters to Ignore...to FollowEndorsed Only
Trim Comments?
    []
  1. res says:

    Thanks to you and Piffer for providing this!

    I just downloaded the OSF files. GWAS_EA.to10K.txt appears to be missing. I assume it is the file at https://www.thessgac.org/data so downloaded from there.

    The file.choose mechanism is inconvenient for repeated runs. Would it be possible (I could do this if wanted) to include a small function which takes a binary flag (defined in a single place at the start of the file) and either calls file.choose or loads the default file? Here is a simple version if anyone wants it.

    choose.file <- function(filename, use.chooser=FALSE) {
    if (use.chooser) {
    file.choose()
    } else {
    filename
    }
    }

    use.chooser <- TRUE

    HGDP_CEPH=read.csv(choose.file("Lee_results_10k_final.csv", use.chooser), header=TRUE, sep = ";")#open HGDP-CEPH browser output with freqs from GWAS hits

    It might help readability (e.g. the second paragraph) to force line breaks in the output where desired. Adding two spaces at the end of the line does this in R Markdown (otherwise it merges successive lines).

    I found the first bar chart easier to read with “fig.height=8″ set.

    The other PGS results and correlations (see either the Rmd files or RPubs) are also very interesting.

    Read More
    • Replies: @FKA Max
    res,

    could you, please, do me a favor and calculate the correlation between the population/nation Met allele frequencies of Mr. Piffer's following paper: https://lesacreduprintemps19.files.wordpress.com/2014/01/correlation-of-the-comt-val158met-polymorphism-with-latitude-and-a-hunter-gather-lifestyle-suggests-culturee28093gene-coevolution-and-selective-pressure-on-cognition-genes-due-to-climate.pdf

    https://2kpcwh2r7phz1nq4jj237m22-wpengine.netdna-ssl.com/wp-content/uploads/2014/01/MetFrequency.jpg

    and Afrosapiens's IQ data instead of Lynn's IQ data, because I believe the correlation will be even stronger and will explain Piffer's contradictory finding: "Particularly interesting is the relatively low frequency of COMT in East Asian populations (range 0.22–0.30), which contrasts with their reported higher IQ (105)." - p. 169

    https://notpolitcallycorrect.files.wordpress.com/2017/09/ranking.png

    Source: https://notpoliticallycorrect.me/2017/09/05/worldwide-iq-estimates-based-on-education-data/

    This is what I wrote a couple of months ago on this:

    If East Asians were not such hard-working students http://www.unz.com/freed/fun-with-iq-deep-thought/#comment-2095195 and were not performing so well on standardized IQ tests due to their studying habits th[e]n the global correlation between IQ and population COMT Met frequencies would be even higher, in my opinion:

    Correlation of the COMT Val158Met Polymorphism with Latitude and a Hunter-Gather Lifestyle Suggests Culture–Gene Coevolution and Selective Pressure on Cognition Genes Due to Climate

    Davide Piffer, Anthropological Science, July 31, 2013

    Thus, the global correlation between IQ and Met allele frequency is r = 0.579 and highly statistically significant (n = 38; P < 0.001). This supports the prediction that populations with higher Met allele frequency have higher IQ, similarly to the correlation observed at the individual level.

    – https://www.amren.com/news/2014/01/correlation-of-the-comt-val158met-polymorphism-with-latitude-and-a-hunter-gather-lifestyle-suggests-culture-gene-coevolution-and-selective-pressure-on-cognition-genes-due-to-climate/
     

    - http://www.unz.com/jderbyshire/time-to-stop-importing-an-immigrant-overclass/#comment-2118604

    I agree with Afrosapiens that Lynn has overestimated East Asian intelligence, and I believe that COMT Met is still the best genetic predictor of intelligence, even though some researchers like Emil Kirkegaard disagree:

    Just commented on this issue over in another Unz Review comments thread, Mr. Thompson, and how these emotional and genetic differences can even affect IQ test scores, etc., since stress-susceptibility is quite a significant moderating/confounding factor when it comes to stressful test taking conditions, in my opinion.
    [...]
    East Asians and Africans, on average, have a competitive advantage under stressful test taking conditions over Caucasians and Mexicans, etc. due to this, in my opinion.
    [...]
    Mr. Kirkegaard thinks that these are “More failed candidate gene ideas.”, but I believe he is jumping to premature conclusion, due to the reasons I stated above
     

    - https://www.unz.com/jthompson/the-anatomy-of-melancholy/#comment-2124173

    In this video, the person https://en.wikipedia.org/wiki/Po_Bronson interviewed claims/estimates that “worriers/strategists” have a 10 IQ point advantage over “warriors” in non-stressful situations/environments. - https://www.unz.com/jthompson/the-anatomy-of-melancholy/#comment-2124219

    https://www.youtube.com/watch?v=S2_3RXmXoM8

    Thank you very much, res.

    ReplyAgree/Disagree/Etc.
    AgreeDisagreeLOLTroll
    These buttons register your public Agreement, Disagreement, Troll, or LOL with the selected comment. They are ONLY available to recent, frequent commenters who have saved their Name+Email using the 'Remember My Information' checkbox, and may also ONLY be used once per hour.
    Ignore Commenter Follow Commenter
    More... This Commenter This Thread Hide Thread Display All Comments
  2. j2 says:

    Happy to see that you are a bit skeptical of Piffer’s results, I thought I was the only one. They are very good and quite what you might want, and that is the problem I have with them. The polygenic score approach itself is a question mark. Most mutations should be recessive, yet in the polygenic score you count 1=heterozygote, 2=homozygote, as if it was partially dominant. I have a somewhat different view of intelligence, wrote a post here

    http://www.pienisalaliittotutkimus.com/2018/04/20/men-inherit-male-intelligence-from-mom-but-not-the-iq-variance/

    So, I think men have from the X chromosome a different brain structure, and additionally there are the IQ-boosting autosomal genes and the IQ-lowering recessive X-linked genes, that in men get expressed. Piffer’s papers simplify a difficult issue too much to my liking and the results are better than I would expect.

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter Display All Comments
  3. dearieme says:

    “merciless enquiry and savage criticism”: you mean, is Piffer piffle?

    Read More
    • Replies: @James Thompson
    precisely. a souffle of barbed wire.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  4. FKA Max says: • Website
    @res
    Thanks to you and Piffer for providing this!

    I just downloaded the OSF files. GWAS_EA.to10K.txt appears to be missing. I assume it is the file at https://www.thessgac.org/data so downloaded from there.

    The file.choose mechanism is inconvenient for repeated runs. Would it be possible (I could do this if wanted) to include a small function which takes a binary flag (defined in a single place at the start of the file) and either calls file.choose or loads the default file? Here is a simple version if anyone wants it.

    choose.file <- function(filename, use.chooser=FALSE) {
    if (use.chooser) {
    file.choose()
    } else {
    filename
    }
    }

    use.chooser <- TRUE

    HGDP_CEPH=read.csv(choose.file("Lee_results_10k_final.csv", use.chooser), header=TRUE, sep = ";")#open HGDP-CEPH browser output with freqs from GWAS hits
     
    It might help readability (e.g. the second paragraph) to force line breaks in the output where desired. Adding two spaces at the end of the line does this in R Markdown (otherwise it merges successive lines).

    I found the first bar chart easier to read with "fig.height=8" set.

    The other PGS results and correlations (see either the Rmd files or RPubs) are also very interesting.

    res,

    could you, please, do me a favor and calculate the correlation between the population/nation Met allele frequencies of Mr. Piffer’s following paper: https://lesacreduprintemps19.files.wordpress.com/2014/01/correlation-of-the-comt-val158met-polymorphism-with-latitude-and-a-hunter-gather-lifestyle-suggests-culturee28093gene-coevolution-and-selective-pressure-on-cognition-genes-due-to-climate.pdf

    and Afrosapiens‘s IQ data instead of Lynn’s IQ data, because I believe the correlation will be even stronger and will explain Piffer’s contradictory finding: “Particularly interesting is the relatively low frequency of COMT in East Asian populations (range 0.22–0.30), which contrasts with their reported higher IQ (105).” – p. 169

    Source: https://notpoliticallycorrect.me/2017/09/05/worldwide-iq-estimates-based-on-education-data/

    This is what I wrote a couple of months ago on this:

    If East Asians were not such hard-working students http://www.unz.com/freed/fun-with-iq-deep-thought/#comment-2095195 and were not performing so well on standardized IQ tests due to their studying habits th[e]n the global correlation between IQ and population COMT Met frequencies would be even higher, in my opinion:

    Correlation of the COMT Val158Met Polymorphism with Latitude and a Hunter-Gather Lifestyle Suggests Culture–Gene Coevolution and Selective Pressure on Cognition Genes Due to Climate

    Davide Piffer, Anthropological Science, July 31, 2013

    Thus, the global correlation between IQ and Met allele frequency is r = 0.579 and highly statistically significant (n = 38; P < 0.001). This supports the prediction that populations with higher Met allele frequency have higher IQ, similarly to the correlation observed at the individual level.

    https://www.amren.com/news/2014/01/correlation-of-the-comt-val158met-polymorphism-with-latitude-and-a-hunter-gather-lifestyle-suggests-culture-gene-coevolution-and-selective-pressure-on-cognition-genes-due-to-climate/

    http://www.unz.com/jderbyshire/time-to-stop-importing-an-immigrant-overclass/#comment-2118604

    I agree with Afrosapiens that Lynn has overestimated East Asian intelligence, and I believe that COMT Met is still the best genetic predictor of intelligence, even though some researchers like Emil Kirkegaard disagree:

    Just commented on this issue over in another Unz Review comments thread, Mr. Thompson, and how these emotional and genetic differences can even affect IQ test scores, etc., since stress-susceptibility is quite a significant moderating/confounding factor when it comes to stressful test taking conditions, in my opinion.
    [...]
    East Asians and Africans, on average, have a competitive advantage under stressful test taking conditions over Caucasians and Mexicans, etc. due to this, in my opinion.
    [...]
    Mr. Kirkegaard thinks that these are “More failed candidate gene ideas.”, but I believe he is jumping to premature conclusion, due to the reasons I stated above

    https://www.unz.com/jthompson/the-anatomy-of-melancholy/#comment-2124173

    In this video, the person https://en.wikipedia.org/wiki/Po_Bronson interviewed claims/estimates that “worriers/strategists” have a 10 IQ point advantage over “warriors” in non-stressful situations/environments. https://www.unz.com/jthompson/the-anatomy-of-melancholy/#comment-2124219

    Thank you very much, res.

    Read More
    • Replies: @res
    Sounds like something worth testing. I think the best way to do this is to incorporate the additional data into the analysis in http://rpubs.com/Daxide/377423
    I have that running locally so all I need is the appropriate data.

    The easiest way to do that would be to modify the HGDP_PGS .csv available at https://osf.io/uays8/
    Note the specific populations needed in the first column.

    Would it be possible for you to create a file adding the fields you want to that? I assume you would be looking at Lynn IQ, Afro IQ, and % Met?

    I can create the data file, but there are enough population mapping issues that I would prefer the person interested in the results make those assumptions.

    P.S. If Davide would like to look at this himself IMHO that would be even better.

    P.P.S. Do you have any thoughts why rs4680 does not show up in the results for the latest EA study? https://www.snpedia.com/index.php/Rs4680
    Though that page does show an IQ related connection: https://www.ncbi.nlm.nih.gov/pubmed/24853458?dopt=Abstract
    There is one SNP in the EA study that is in fairly high LD (D' = 0.8, r^2 =0.28 in CEU) with rs4680: rs2240715, but the p value is 0.7 with a small beta (I did not check exhaustively, just a quick visual look at chr 22 hits). Actually, rs165655 is also similar.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  5. res says:
    @FKA Max
    res,

    could you, please, do me a favor and calculate the correlation between the population/nation Met allele frequencies of Mr. Piffer's following paper: https://lesacreduprintemps19.files.wordpress.com/2014/01/correlation-of-the-comt-val158met-polymorphism-with-latitude-and-a-hunter-gather-lifestyle-suggests-culturee28093gene-coevolution-and-selective-pressure-on-cognition-genes-due-to-climate.pdf

    https://2kpcwh2r7phz1nq4jj237m22-wpengine.netdna-ssl.com/wp-content/uploads/2014/01/MetFrequency.jpg

    and Afrosapiens's IQ data instead of Lynn's IQ data, because I believe the correlation will be even stronger and will explain Piffer's contradictory finding: "Particularly interesting is the relatively low frequency of COMT in East Asian populations (range 0.22–0.30), which contrasts with their reported higher IQ (105)." - p. 169

    https://notpolitcallycorrect.files.wordpress.com/2017/09/ranking.png

    Source: https://notpoliticallycorrect.me/2017/09/05/worldwide-iq-estimates-based-on-education-data/

    This is what I wrote a couple of months ago on this:

    If East Asians were not such hard-working students http://www.unz.com/freed/fun-with-iq-deep-thought/#comment-2095195 and were not performing so well on standardized IQ tests due to their studying habits th[e]n the global correlation between IQ and population COMT Met frequencies would be even higher, in my opinion:

    Correlation of the COMT Val158Met Polymorphism with Latitude and a Hunter-Gather Lifestyle Suggests Culture–Gene Coevolution and Selective Pressure on Cognition Genes Due to Climate

    Davide Piffer, Anthropological Science, July 31, 2013

    Thus, the global correlation between IQ and Met allele frequency is r = 0.579 and highly statistically significant (n = 38; P < 0.001). This supports the prediction that populations with higher Met allele frequency have higher IQ, similarly to the correlation observed at the individual level.

    – https://www.amren.com/news/2014/01/correlation-of-the-comt-val158met-polymorphism-with-latitude-and-a-hunter-gather-lifestyle-suggests-culture-gene-coevolution-and-selective-pressure-on-cognition-genes-due-to-climate/
     

    - http://www.unz.com/jderbyshire/time-to-stop-importing-an-immigrant-overclass/#comment-2118604

    I agree with Afrosapiens that Lynn has overestimated East Asian intelligence, and I believe that COMT Met is still the best genetic predictor of intelligence, even though some researchers like Emil Kirkegaard disagree:

    Just commented on this issue over in another Unz Review comments thread, Mr. Thompson, and how these emotional and genetic differences can even affect IQ test scores, etc., since stress-susceptibility is quite a significant moderating/confounding factor when it comes to stressful test taking conditions, in my opinion.
    [...]
    East Asians and Africans, on average, have a competitive advantage under stressful test taking conditions over Caucasians and Mexicans, etc. due to this, in my opinion.
    [...]
    Mr. Kirkegaard thinks that these are “More failed candidate gene ideas.”, but I believe he is jumping to premature conclusion, due to the reasons I stated above
     

    - https://www.unz.com/jthompson/the-anatomy-of-melancholy/#comment-2124173

    In this video, the person https://en.wikipedia.org/wiki/Po_Bronson interviewed claims/estimates that “worriers/strategists” have a 10 IQ point advantage over “warriors” in non-stressful situations/environments. - https://www.unz.com/jthompson/the-anatomy-of-melancholy/#comment-2124219

    https://www.youtube.com/watch?v=S2_3RXmXoM8

    Thank you very much, res.

    Sounds like something worth testing. I think the best way to do this is to incorporate the additional data into the analysis in http://rpubs.com/Daxide/377423
    I have that running locally so all I need is the appropriate data.

    The easiest way to do that would be to modify the HGDP_PGS .csv available at https://osf.io/uays8/
    Note the specific populations needed in the first column.

    Would it be possible for you to create a file adding the fields you want to that? I assume you would be looking at Lynn IQ, Afro IQ, and % Met?

    I can create the data file, but there are enough population mapping issues that I would prefer the person interested in the results make those assumptions.

    P.S. If Davide would like to look at this himself IMHO that would be even better.

    P.P.S. Do you have any thoughts why rs4680 does not show up in the results for the latest EA study? https://www.snpedia.com/index.php/Rs4680
    Though that page does show an IQ related connection: https://www.ncbi.nlm.nih.gov/pubmed/24853458?dopt=Abstract
    There is one SNP in the EA study that is in fairly high LD (D’ = 0.8, r^2 =0.28 in CEU) with rs4680: rs2240715, but the p value is 0.7 with a small beta (I did not check exhaustively, just a quick visual look at chr 22 hits). Actually, rs165655 is also similar.

    Read More
    • Replies: @FKA Max
    Thanks so much, res.

    I used an online correlation coefficient and linear regression calculator and calculated the correlation with Afrosapiens IQ data ( https://notpoliticallycorrect.me/2017/09/05/worldwide-iq-estimates-based-on-education-data/ see "Results") for thirty-two (32) populations/nations:

    Sample size: 32

    Correlation coefficient (r): 0.41594921999427 : http://www.alcula.com/calculators/statistics/correlation-coefficient/

    Sample size: 32
    Mean x (x̄): 86.75
    Mean y (ȳ): 0.398859375
    Intercept (a): -0.037613843077569
    Slope (b): 0.0050313915628538
    Regression line equation: y=0.0050313915628538x-0.037613843077569 : http://www.alcula.com/calculators/statistics/correlation-coefficient/

    Image linear regression: http://www.alcula.com/calculators/statistics/linear-regression/linear-regression-image.php?n=1

    IQ,Met% Country
    104,0.609 Denmark
    102,0.512 Ireland
    100,0.53 UK
    99,0.572 Estonia
    99,0.248 Korea
    97,0.529 Finland
    97,0.47 France
    97,0.304 Japan
    96,0.533 Hungary
    96,0.46 Italy
    96,0.43 Spain
    95,0.478 Russia
    91,0.25 Mongolia
    89,0.5 Iran
    87,0.45 Turkey
    86,0.416 Lebanon
    86,0.6 Mexico
    85,0.221 Thailand
    85,0.57 Palestine
    85,0.135 Micronesia
    84,0.243 China
    83,0.3 Vietnam
    79,0.232 India
    79,0.2475 Ghana
    77,0.345 Kenya
    75,0.332 Nigeria
    74,0.297 Cambodia
    71,0.416 Syria
    72,0.44 Papa New Guinea
    72,0.299 Tanzania
    70,0.465 Pakistan
    68,0.33 Senegal

    It actually weakened the correlation from r = 0.579 Lynn to r = 0.415 Afrosapiens.
    , @dux.ie
    Re: P.P.S. Do you have any thoughts why rs4680 does not show up in the results for the latest EA study? https://www.snpedia.com/index.php/Rs4680

    I have been saying this for a few years now. If you read the snpedia article closely, for warrior gene it talks of "higher pain threshold, better stress resiliency (in work places and group competitions), albeit with a modest reduction in executive cognition performance under most conditions (in home and individual undertakings)", the warrior gene is more like a 'grit gene' than a 'IQ gene'. The effects of rs4680 on IQ seems to be secondary. The performance of person with high grit without accompany high IQ will only be modest. The effects of 'performance anxiety' under group competitions can be seen in the OECD PISA data.

    The PISA project have data on the percentage of students who wanted to be the best, the higher that percentage the higher the competitive stress or anxiety, the results is shown in the chart,

    http://tinypic.com/view.php?pic=2v8fztg&s=9

    It can be seen that on average overall competitive stress reduces performance. However in residual analysis there is a dense cluster of outliers indicating that there are finer grain relationship which opposes the overall downward trend, i.e. the effects of stress which might be influenced by rs4680 is non-linear and exibiting a pitch-fork shape. The finer grain trend can be determined objectively and simply by dividing the data into 4 equal size quadrants and the data in the first quadrant gives the opposing trend.It is interesting that the countries in the first quadrant are the EastAsians, the Scandinavian countries from the descendents of the Vikings, the recently frontier countries or home countries of the descendents of the Anglo-Saxon/Celtic/Viking. The odd country was Japan whose students were not competitive and that had been identified in various studies as due to the national angst from losing WW2 that there could be epigenetic changes to the genomes the effects of which might last a few generations.

    The chart of PISA scores against the percentage of population percentage of rs4680 variant,

    http://tinypic.com/view.php?pic=nv9ph2&s=9

    If there is no overall statistical trends but simple clustering of the data space into two equal sets can produce two significant opposing trends, the hypothesis that there is no statistically significant result should be rejected. Note the rs4680 data for some countries in the lower fork with low PISA3 scores and most probably high FreqG are not available and the results showed a V shape. High performance can be achieved in both extreme high and low FreqG (or FreqA), thus it is primarily not an IQ gene.

    GWAS with the one track mind of only considering linear trends, the non-linear trends can easily be omitted.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  6. @dearieme
    "merciless enquiry and savage criticism": you mean, is Piffer piffle?

    precisely. a souffle of barbed wire.

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  7. FKA Max says: • Website
    @res
    Sounds like something worth testing. I think the best way to do this is to incorporate the additional data into the analysis in http://rpubs.com/Daxide/377423
    I have that running locally so all I need is the appropriate data.

    The easiest way to do that would be to modify the HGDP_PGS .csv available at https://osf.io/uays8/
    Note the specific populations needed in the first column.

    Would it be possible for you to create a file adding the fields you want to that? I assume you would be looking at Lynn IQ, Afro IQ, and % Met?

    I can create the data file, but there are enough population mapping issues that I would prefer the person interested in the results make those assumptions.

    P.S. If Davide would like to look at this himself IMHO that would be even better.

    P.P.S. Do you have any thoughts why rs4680 does not show up in the results for the latest EA study? https://www.snpedia.com/index.php/Rs4680
    Though that page does show an IQ related connection: https://www.ncbi.nlm.nih.gov/pubmed/24853458?dopt=Abstract
    There is one SNP in the EA study that is in fairly high LD (D' = 0.8, r^2 =0.28 in CEU) with rs4680: rs2240715, but the p value is 0.7 with a small beta (I did not check exhaustively, just a quick visual look at chr 22 hits). Actually, rs165655 is also similar.

    Thanks so much, res.

    I used an online correlation coefficient and linear regression calculator and calculated the correlation with Afrosapiens IQ data ( https://notpoliticallycorrect.me/2017/09/05/worldwide-iq-estimates-based-on-education-data/ see “Results”) for thirty-two (32) populations/nations:

    Sample size: 32

    Correlation coefficient (r): 0.41594921999427 : http://www.alcula.com/calculators/statistics/correlation-coefficient/

    Sample size: 32
    Mean x (x̄): 86.75
    Mean y (ȳ): 0.398859375
    Intercept (a): -0.037613843077569
    Slope (b): 0.0050313915628538
    Regression line equation: y=0.0050313915628538x-0.037613843077569 : http://www.alcula.com/calculators/statistics/correlation-coefficient/

    Image linear regression: http://www.alcula.com/calculators/statistics/linear-regression/linear-regression-image.php?n=1

    [MORE]

    IQ,Met% Country
    104,0.609 Denmark
    102,0.512 Ireland
    100,0.53 UK
    99,0.572 Estonia
    99,0.248 Korea
    97,0.529 Finland
    97,0.47 France
    97,0.304 Japan
    96,0.533 Hungary
    96,0.46 Italy
    96,0.43 Spain
    95,0.478 Russia
    91,0.25 Mongolia
    89,0.5 Iran
    87,0.45 Turkey
    86,0.416 Lebanon
    86,0.6 Mexico
    85,0.221 Thailand
    85,0.57 Palestine
    85,0.135 Micronesia
    84,0.243 China
    83,0.3 Vietnam
    79,0.232 India
    79,0.2475 Ghana
    77,0.345 Kenya
    75,0.332 Nigeria
    74,0.297 Cambodia
    71,0.416 Syria
    72,0.44 Papa New Guinea
    72,0.299 Tanzania
    70,0.465 Pakistan
    68,0.33 Senegal

    It actually weakened the correlation from r = 0.579 Lynn to r = 0.415 Afrosapiens.

    Read More
    • Replies: @FKA Max
    Correction: I shared the wrong link for the linear regression calculator I used, here the correct one: http://www.alcula.com/calculators/statistics/linear-regression/

    And again, here the Piffer Met% data I used: https://www.amren.com/news/2014/01/correlation-of-the-comt-val158met-polymorphism-with-latitude-and-a-hunter-gather-lifestyle-suggests-culture-gene-coevolution-and-selective-pressure-on-cognition-genes-due-to-climate/

    , @res
    Thanks. I downloaded the data for rs4680 from Piffer's source at http://spsmart.cesga.es/ceph.php?dataSet=ceph_stanford
    which allows me to look at COMT Met% for the exact populations Piffer used. Data below.

    I'll probably try to map the Lynn and Afrosapiens IQ estimates into Piffer's populations as I described in my earlier comment (unless you do it first). I am a bit surprised he did not do the Lynn comparison himself in the correlation page I linked.


    It actually weakened the correlation from r = 0.579 Lynn to r = 0.415 Afrosapiens.
     
    This does not surprise me. Although Afrosapiens's technique is interesting, using it for between country IQ comparisons of countries with widely differing levels of development has an obvious issue with the EA-development correlation interfering with the EA-IQ correlation. I think it is much better for within country comparisons (as in a typical EA GWAS) or countries with similar levels of development. Which makes it ironic that Afrosapiens touts his method in the worldwide IQ comparison context but was strangely silent when I brought it up to argue for an EA-IQ link in Nigeria (in a Chanda Chisala thread). That exchange helped make clear just how selective and motivated his reasoning is.

    P.S. Some links to the Nigeria EA-IQ conversation: http://www.unz.com/article/my-last-word-on-the-scrabble-and-iq-debate-2/#comment-2016734
    http://www.unz.com/article/my-last-word-on-the-scrabble-and-iq-debate-2/#comment-2012772
    I still can't believe Chanda tried to argue for a zero or even negative correlation between EA and IQ in Nigeria... SMH

    P.P.S. COMT Met% (rs4680 A allele) for Piffer populations below.



    Continent level:

    population N freq_A
    Population Set 1 944 0.389
    AFRICA 102 0.275
    AMERICA 64 0.305
    EUROPE 158 0.494
    MIDDLE EAST 163 0.463
    CENTRAL-SOUTH ASIA 200 0.477
    OCEANIA 28 0.393
    EAST ASIA 229 0.26

    These continent level numbers (particularly East Asia) make me skeptical about a strong Met% IQ connection.

    Subpopulation level:

    population N freq_A
    Population Set 1 944 0.389
    C. African Republic - Biaka Pygmy 22 0.068
    D. R. of Congo - Mbuti Pygmy 13 0.231
    Kenya - Bantu 11 0.364
    Namibia - San 5 0.1
    Nigeria - Yoruba 21 0.405
    Senegal - Mandenka 22 0.364
    South Africa - Bantu 8 0.313
    Brazil - Karitiana 14 0.036
    Brazil - Surui 8 0.313
    Colombia - Piapoco and Curripaco 7 0
    Mexico - Maya 21 0.571
    Mexico - Pima 14 0.321
    France - Basque 24 0.583
    France - French 28 0.482
    Italy - Sardinian 28 0.357
    Italy - Tuscan 8 0.5
    Italy - from Bergamo 13 0.423
    Orkney Islands - Orcadian 15 0.6
    Russia (Caucasus) - Adygei 17 0.412
    Russia - Russian 25 0.6
    Algeria (Mzab) - Mozabite 29 0.431
    Israel (Carmel) - Druze 42 0.464
    Israel (Central) - Palestinian 46 0.533
    Israel (Negev) - Bedouin 46 0.413
    China - Uygur 10 0.45
    Pakistan - Balochi 24 0.542
    Pakistan - Brahui 25 0.42
    Pakistan - Burusho 25 0.48
    Pakistan - Hazara 22 0.523
    Pakistan - Kalash 23 0.565
    Pakistan - Makrani 25 0.46
    Pakistan - Pathan 22 0.432
    Pakistan - Sindhi 24 0.417
    Bougainville - NAN Melanesian 11 0.318
    New Guinea - Papuan 17 0.441
    Cambodia - Cambodian 10 0.35
    China - Dai 10 0.2
    China - Daur 9 0.278
    China - Han 44 0.25
    China - Hezhen 9 0.333
    China - Lahu 8 0.313
    China - Miaozu 10 0.15
    China - Mongola 10 0.25
    China - Naxi 8 0.125
    China - Oroqen 9 0.222
    China - She 10 0.3
    China - Tu 10 0.3
    China - Tujia 10 0.2
    China - Xibo 9 0.167
    China - Yizu 10 0.25
    Japan - Japanese 28 0.304
    Siberia - Yakut 25 0.3
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  8. FKA Max says: • Website
    @FKA Max
    Thanks so much, res.

    I used an online correlation coefficient and linear regression calculator and calculated the correlation with Afrosapiens IQ data ( https://notpoliticallycorrect.me/2017/09/05/worldwide-iq-estimates-based-on-education-data/ see "Results") for thirty-two (32) populations/nations:

    Sample size: 32

    Correlation coefficient (r): 0.41594921999427 : http://www.alcula.com/calculators/statistics/correlation-coefficient/

    Sample size: 32
    Mean x (x̄): 86.75
    Mean y (ȳ): 0.398859375
    Intercept (a): -0.037613843077569
    Slope (b): 0.0050313915628538
    Regression line equation: y=0.0050313915628538x-0.037613843077569 : http://www.alcula.com/calculators/statistics/correlation-coefficient/

    Image linear regression: http://www.alcula.com/calculators/statistics/linear-regression/linear-regression-image.php?n=1

    IQ,Met% Country
    104,0.609 Denmark
    102,0.512 Ireland
    100,0.53 UK
    99,0.572 Estonia
    99,0.248 Korea
    97,0.529 Finland
    97,0.47 France
    97,0.304 Japan
    96,0.533 Hungary
    96,0.46 Italy
    96,0.43 Spain
    95,0.478 Russia
    91,0.25 Mongolia
    89,0.5 Iran
    87,0.45 Turkey
    86,0.416 Lebanon
    86,0.6 Mexico
    85,0.221 Thailand
    85,0.57 Palestine
    85,0.135 Micronesia
    84,0.243 China
    83,0.3 Vietnam
    79,0.232 India
    79,0.2475 Ghana
    77,0.345 Kenya
    75,0.332 Nigeria
    74,0.297 Cambodia
    71,0.416 Syria
    72,0.44 Papa New Guinea
    72,0.299 Tanzania
    70,0.465 Pakistan
    68,0.33 Senegal

    It actually weakened the correlation from r = 0.579 Lynn to r = 0.415 Afrosapiens.
    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  9. res says:
    @FKA Max
    Thanks so much, res.

    I used an online correlation coefficient and linear regression calculator and calculated the correlation with Afrosapiens IQ data ( https://notpoliticallycorrect.me/2017/09/05/worldwide-iq-estimates-based-on-education-data/ see "Results") for thirty-two (32) populations/nations:

    Sample size: 32

    Correlation coefficient (r): 0.41594921999427 : http://www.alcula.com/calculators/statistics/correlation-coefficient/

    Sample size: 32
    Mean x (x̄): 86.75
    Mean y (ȳ): 0.398859375
    Intercept (a): -0.037613843077569
    Slope (b): 0.0050313915628538
    Regression line equation: y=0.0050313915628538x-0.037613843077569 : http://www.alcula.com/calculators/statistics/correlation-coefficient/

    Image linear regression: http://www.alcula.com/calculators/statistics/linear-regression/linear-regression-image.php?n=1

    IQ,Met% Country
    104,0.609 Denmark
    102,0.512 Ireland
    100,0.53 UK
    99,0.572 Estonia
    99,0.248 Korea
    97,0.529 Finland
    97,0.47 France
    97,0.304 Japan
    96,0.533 Hungary
    96,0.46 Italy
    96,0.43 Spain
    95,0.478 Russia
    91,0.25 Mongolia
    89,0.5 Iran
    87,0.45 Turkey
    86,0.416 Lebanon
    86,0.6 Mexico
    85,0.221 Thailand
    85,0.57 Palestine
    85,0.135 Micronesia
    84,0.243 China
    83,0.3 Vietnam
    79,0.232 India
    79,0.2475 Ghana
    77,0.345 Kenya
    75,0.332 Nigeria
    74,0.297 Cambodia
    71,0.416 Syria
    72,0.44 Papa New Guinea
    72,0.299 Tanzania
    70,0.465 Pakistan
    68,0.33 Senegal

    It actually weakened the correlation from r = 0.579 Lynn to r = 0.415 Afrosapiens.

    Thanks. I downloaded the data for rs4680 from Piffer’s source at http://spsmart.cesga.es/ceph.php?dataSet=ceph_stanford
    which allows me to look at COMT Met% for the exact populations Piffer used. Data below.

    I’ll probably try to map the Lynn and Afrosapiens IQ estimates into Piffer’s populations as I described in my earlier comment (unless you do it first). I am a bit surprised he did not do the Lynn comparison himself in the correlation page I linked.

    It actually weakened the correlation from r = 0.579 Lynn to r = 0.415 Afrosapiens.

    This does not surprise me. Although Afrosapiens’s technique is interesting, using it for between country IQ comparisons of countries with widely differing levels of development has an obvious issue with the EA-development correlation interfering with the EA-IQ correlation. I think it is much better for within country comparisons (as in a typical EA GWAS) or countries with similar levels of development. Which makes it ironic that Afrosapiens touts his method in the worldwide IQ comparison context but was strangely silent when I brought it up to argue for an EA-IQ link in Nigeria (in a Chanda Chisala thread). That exchange helped make clear just how selective and motivated his reasoning is.

    P.S. Some links to the Nigeria EA-IQ conversation: http://www.unz.com/article/my-last-word-on-the-scrabble-and-iq-debate-2/#comment-2016734

    http://www.unz.com/article/my-last-word-on-the-scrabble-and-iq-debate-2/#comment-2012772

    I still can’t believe Chanda tried to argue for a zero or even negative correlation between EA and IQ in Nigeria… SMH

    P.P.S. COMT Met% (rs4680 A allele) for Piffer populations below.

    [MORE]

    Continent level:

    population N freq_A
    Population Set 1 944 0.389
    AFRICA 102 0.275
    AMERICA 64 0.305
    EUROPE 158 0.494
    MIDDLE EAST 163 0.463
    CENTRAL-SOUTH ASIA 200 0.477
    OCEANIA 28 0.393
    EAST ASIA 229 0.26

    These continent level numbers (particularly East Asia) make me skeptical about a strong Met% IQ connection.

    Subpopulation level:

    population N freq_A
    Population Set 1 944 0.389
    C. African Republic – Biaka Pygmy 22 0.068
    D. R. of Congo – Mbuti Pygmy 13 0.231
    Kenya – Bantu 11 0.364
    Namibia – San 5 0.1
    Nigeria – Yoruba 21 0.405
    Senegal – Mandenka 22 0.364
    South Africa – Bantu 8 0.313
    Brazil – Karitiana 14 0.036
    Brazil – Surui 8 0.313
    Colombia – Piapoco and Curripaco 7 0
    Mexico – Maya 21 0.571
    Mexico – Pima 14 0.321
    France – Basque 24 0.583
    France – French 28 0.482
    Italy – Sardinian 28 0.357
    Italy – Tuscan 8 0.5
    Italy – from Bergamo 13 0.423
    Orkney Islands – Orcadian 15 0.6
    Russia (Caucasus) – Adygei 17 0.412
    Russia – Russian 25 0.6
    Algeria (Mzab) – Mozabite 29 0.431
    Israel (Carmel) – Druze 42 0.464
    Israel (Central) – Palestinian 46 0.533
    Israel (Negev) – Bedouin 46 0.413
    China – Uygur 10 0.45
    Pakistan – Balochi 24 0.542
    Pakistan – Brahui 25 0.42
    Pakistan – Burusho 25 0.48
    Pakistan – Hazara 22 0.523
    Pakistan – Kalash 23 0.565
    Pakistan – Makrani 25 0.46
    Pakistan – Pathan 22 0.432
    Pakistan – Sindhi 24 0.417
    Bougainville – NAN Melanesian 11 0.318
    New Guinea – Papuan 17 0.441
    Cambodia – Cambodian 10 0.35
    China – Dai 10 0.2
    China – Daur 9 0.278
    China – Han 44 0.25
    China – Hezhen 9 0.333
    China – Lahu 8 0.313
    China – Miaozu 10 0.15
    China – Mongola 10 0.25
    China – Naxi 8 0.125
    China – Oroqen 9 0.222
    China – She 10 0.3
    China – Tu 10 0.3
    China – Tujia 10 0.2
    China – Xibo 9 0.167
    China – Yizu 10 0.25
    Japan – Japanese 28 0.304
    Siberia – Yakut 25 0.3

    Read More
    • Replies: @FKA Max
    Thanks again, res.

    These continent level numbers (particularly East Asia) make me skeptical about a strong Met% IQ connection.
     
    I understand your skepticism.

    This does not surprise me. Although Afrosapiens’s technique is interesting, using it for between country IQ comparisons of countries with widely differing levels of development has an obvious issue with the EA-development correlation interfering with the EA-IQ correlation.
     
    Yes, that is correct. His Syria IQ of 71 is probably due to current the war in Syria, for example. Mexico is also affected by drug wars, etc. The low Pakistani results could be due to high cousin marriage rates, which depress intelligence, high fertility rates and resultant resource scarcity/depletion, religious instead of scientific/academic teaching/study, etc.

    I think the Met% IQ correlation will become clearer and more pronounced the more peaceful, less religious/more secular and the more per-capita prosperous the world and/or a population/group becomes.

    The populations with higher Met% will likely experience "Super-Flynn Effects", when conditions are favorable for them:

    Race/IQ: Super-Flynn Effects in Germans, Jews, and Hispanics

    The central argument of my piece had been that although GDP and IQ were highly correlated, the direction of causality might well be from the former to the latter, and this attracted much derision.
     
    - http://www.unz.com/runz/raceiq-super-flynn-effects-in-germans-jews-and-hispanics/

    Put another way, the populations/groups with lower Met% might have a lower IQ ceiling and potential, and only experience "Semi-Flynn Effects" even under favorable conditions.

    These are some of the reasons why I am skeptical of test results coming out of East Asia, in particular China:


    The National Center for Fair and Open Testing, a nonprofit known as FairTest, which advocates against the misuse and abuse of standardized tests, said Chinese test prep companies have reported cancellations in Taiwan, Macao, Hong Kong, Japan, Singapore and Shanghai.
     
    - https://www.washingtonpost.com/news/answer-sheet/wp/2017/09/07/security-breach-forces-act-test-cancellation-in-asia-and-weather-causes-some-in-u-s/


    But as with just about everything concerning modern China, the results should also be viewed with some distance and possible skepticism. The 5000+ students who were tested in China's biggest and most modern city may or may not be indicative of broader progress throughout the country (as the NYT story points out). Anyone who has had experience with schools and testing in China will want to know more about how these tests were administered, supervised, and scored.
     
    - https://www.theatlantic.com/national/archive/2010/12/on-those-stunning-shanghai-test-scores/67654/
    , @FKA Max


    These continent level numbers (particularly East Asia) make me skeptical about a strong Met% IQ connection.
     
    res,

    I found another study that could shed some light on what we were discussing, and actually, I believe, proves both of us right to some extent:

    Association study between COMT 158Met and creativity scores in bipolar disorder and healthy controls

    There are many difficulties inherent to systematic studies of creativity, particularly methodological problems concerning the reliability and validity of creativity measures, and disagreements over the definition of creativity. The BWAS is not the only measure of creativity and other measures should be explored before drawing more definitive conclusions. In the present study, it was decided to examine the correlation between scores on a widely used scale for measuring creativity and the presence of functional polymorphism of COMT (rs4680), which likely influences PFC cognition, in a homogeneous sample of university students. Our results are also consistent with those reported in the literature investigating the role of DA and COMT in PFC function and cognition. However, no influence of COMT on IQ was evident, and BWAS and IQ scores were unrelated, further suggesting some degree of specificity in the association of COMT with creativity.

    This study is the first to report findings that suggest the effects of COMT gene polymorphism may not be limited to isolated basal cognitive abilities, but could partially account for greater cognitive abilities related to creativity in healthy controls.
     
    - http://www.scielo.br/scielo.php?script=sci_arttext&pid=S0101-60832014000200029

    Another Half Brick of Creativity


    The paper is slightly unusual, in that it seeks to model creative processes using IQ120 as a cut-off, and so runs contrary to the general findings of the Lubinski and Benbow work that there is no cut-off point, and that the brighter you are the more creative you are in real life.
     
    - http://www.unz.com/jthompson/another-half-brick-of-creativity/#p_1_9

    Heave Half a Brick at Creativity

    http://lh6.ggpht.com/-6RIi9RaWz0I/VKWLiLxk8fI/AAAAAAAAAuA/LpL_OgIuaec/image_thumb%25255B7%25255D.png

    Source: https://www.unz.com/jthompson/heave-half-brick-at-creativity/

    This raises several questions for me, specifically whether creativity is a better measure of intelligence/brightness than IQ is?

    Population COMT Met frequencies might be the best indicator we have to determine creativity, and lower COMT Met frequencies in East Asian populations might also explain the "low-creativity" stereotype often associated with East Asians.

    ‘Why Do Chinese Lack Creativity?’

    It’s not for lack of trying. The Chinese government now pours billions of dollars annually into research and development — by one estimate, its research and development budget may surpass U.S. spending by 2019 — and Chinese President Xi Jinping has emphasized innovation in his speeches. For the past four years, China has filed more patent applications than any other country, although state news agency Xinhua has described the quality of those patents as “poor.” - http://foreignpolicy.com/2015/06/23/china-innovation-creativity-research-patents/

    I discussed something similar a while back with Afrosapiens:

    Following is my position on IQ tests. Maybe instead of “IQ test”, it should more accurately be called a “ Formal Education Potential, Quality and Attainment test” or something like that. I still feel the tests are useful and can offer some interesting insights, but they surely are and should not be the be-all and end-all tool of how we organize and structure our nations and societies, IMHO:

    I think what is important to reiterate is that IQ tests seem to measure and predict certain things very accurately, e.g., better cognitive performance/functioning under pressure/stress, educational attainment, income, possibly testosterone and dopamine levels, etc., but they do ironically/paradoxically only seem to test “intelligence” to a limited extent, at least that is my best, current understanding and interpretation of the data I have researched thus far.
     
    - http://www.unz.com/jthompson/the-worlds-iq-86/#comment-2072872

    To conclude:

    What are Piffer et al.'s SNPs actually measuring?
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  10. FKA Max says: • Website
    @res
    Thanks. I downloaded the data for rs4680 from Piffer's source at http://spsmart.cesga.es/ceph.php?dataSet=ceph_stanford
    which allows me to look at COMT Met% for the exact populations Piffer used. Data below.

    I'll probably try to map the Lynn and Afrosapiens IQ estimates into Piffer's populations as I described in my earlier comment (unless you do it first). I am a bit surprised he did not do the Lynn comparison himself in the correlation page I linked.


    It actually weakened the correlation from r = 0.579 Lynn to r = 0.415 Afrosapiens.
     
    This does not surprise me. Although Afrosapiens's technique is interesting, using it for between country IQ comparisons of countries with widely differing levels of development has an obvious issue with the EA-development correlation interfering with the EA-IQ correlation. I think it is much better for within country comparisons (as in a typical EA GWAS) or countries with similar levels of development. Which makes it ironic that Afrosapiens touts his method in the worldwide IQ comparison context but was strangely silent when I brought it up to argue for an EA-IQ link in Nigeria (in a Chanda Chisala thread). That exchange helped make clear just how selective and motivated his reasoning is.

    P.S. Some links to the Nigeria EA-IQ conversation: http://www.unz.com/article/my-last-word-on-the-scrabble-and-iq-debate-2/#comment-2016734
    http://www.unz.com/article/my-last-word-on-the-scrabble-and-iq-debate-2/#comment-2012772
    I still can't believe Chanda tried to argue for a zero or even negative correlation between EA and IQ in Nigeria... SMH

    P.P.S. COMT Met% (rs4680 A allele) for Piffer populations below.



    Continent level:

    population N freq_A
    Population Set 1 944 0.389
    AFRICA 102 0.275
    AMERICA 64 0.305
    EUROPE 158 0.494
    MIDDLE EAST 163 0.463
    CENTRAL-SOUTH ASIA 200 0.477
    OCEANIA 28 0.393
    EAST ASIA 229 0.26

    These continent level numbers (particularly East Asia) make me skeptical about a strong Met% IQ connection.

    Subpopulation level:

    population N freq_A
    Population Set 1 944 0.389
    C. African Republic - Biaka Pygmy 22 0.068
    D. R. of Congo - Mbuti Pygmy 13 0.231
    Kenya - Bantu 11 0.364
    Namibia - San 5 0.1
    Nigeria - Yoruba 21 0.405
    Senegal - Mandenka 22 0.364
    South Africa - Bantu 8 0.313
    Brazil - Karitiana 14 0.036
    Brazil - Surui 8 0.313
    Colombia - Piapoco and Curripaco 7 0
    Mexico - Maya 21 0.571
    Mexico - Pima 14 0.321
    France - Basque 24 0.583
    France - French 28 0.482
    Italy - Sardinian 28 0.357
    Italy - Tuscan 8 0.5
    Italy - from Bergamo 13 0.423
    Orkney Islands - Orcadian 15 0.6
    Russia (Caucasus) - Adygei 17 0.412
    Russia - Russian 25 0.6
    Algeria (Mzab) - Mozabite 29 0.431
    Israel (Carmel) - Druze 42 0.464
    Israel (Central) - Palestinian 46 0.533
    Israel (Negev) - Bedouin 46 0.413
    China - Uygur 10 0.45
    Pakistan - Balochi 24 0.542
    Pakistan - Brahui 25 0.42
    Pakistan - Burusho 25 0.48
    Pakistan - Hazara 22 0.523
    Pakistan - Kalash 23 0.565
    Pakistan - Makrani 25 0.46
    Pakistan - Pathan 22 0.432
    Pakistan - Sindhi 24 0.417
    Bougainville - NAN Melanesian 11 0.318
    New Guinea - Papuan 17 0.441
    Cambodia - Cambodian 10 0.35
    China - Dai 10 0.2
    China - Daur 9 0.278
    China - Han 44 0.25
    China - Hezhen 9 0.333
    China - Lahu 8 0.313
    China - Miaozu 10 0.15
    China - Mongola 10 0.25
    China - Naxi 8 0.125
    China - Oroqen 9 0.222
    China - She 10 0.3
    China - Tu 10 0.3
    China - Tujia 10 0.2
    China - Xibo 9 0.167
    China - Yizu 10 0.25
    Japan - Japanese 28 0.304
    Siberia - Yakut 25 0.3

    Thanks again, res.

    These continent level numbers (particularly East Asia) make me skeptical about a strong Met% IQ connection.

    I understand your skepticism.

    This does not surprise me. Although Afrosapiens’s technique is interesting, using it for between country IQ comparisons of countries with widely differing levels of development has an obvious issue with the EA-development correlation interfering with the EA-IQ correlation.

    Yes, that is correct. His Syria IQ of 71 is probably due to current the war in Syria, for example. Mexico is also affected by drug wars, etc. The low Pakistani results could be due to high cousin marriage rates, which depress intelligence, high fertility rates and resultant resource scarcity/depletion, religious instead of scientific/academic teaching/study, etc.

    I think the Met% IQ correlation will become clearer and more pronounced the more peaceful, less religious/more secular and the more per-capita prosperous the world and/or a population/group becomes.

    The populations with higher Met% will likely experience “Super-Flynn Effects”, when conditions are favorable for them:

    Race/IQ: Super-Flynn Effects in Germans, Jews, and Hispanics

    The central argument of my piece had been that although GDP and IQ were highly correlated, the direction of causality might well be from the former to the latter, and this attracted much derision.

    http://www.unz.com/runz/raceiq-super-flynn-effects-in-germans-jews-and-hispanics/

    Put another way, the populations/groups with lower Met% might have a lower IQ ceiling and potential, and only experience “Semi-Flynn Effects” even under favorable conditions.

    These are some of the reasons why I am skeptical of test results coming out of East Asia, in particular China:

    The National Center for Fair and Open Testing, a nonprofit known as FairTest, which advocates against the misuse and abuse of standardized tests, said Chinese test prep companies have reported cancellations in Taiwan, Macao, Hong Kong, Japan, Singapore and Shanghai.

    https://www.washingtonpost.com/news/answer-sheet/wp/2017/09/07/security-breach-forces-act-test-cancellation-in-asia-and-weather-causes-some-in-u-s/

    But as with just about everything concerning modern China, the results should also be viewed with some distance and possible skepticism. The 5000+ students who were tested in China’s biggest and most modern city may or may not be indicative of broader progress throughout the country (as the NYT story points out). Anyone who has had experience with schools and testing in China will want to know more about how these tests were administered, supervised, and scored.

    https://www.theatlantic.com/national/archive/2010/12/on-those-stunning-shanghai-test-scores/67654/

    Read More
    • Replies: @FKA Max
    When I slightly alter/manipulate the sample by taking 85,0.135 Micronesia and 72,0.44 Papa New Guinea out of the sample, changing Syria from IQ 71 to Lebanon's IQ 86 and Pakistan's IQ 70 to India's IQ 79, and Mexico from Met% 0.6 to Met% 0.43 like Spain, I get a correlation coefficient of:

    Sample size: 30

    Correlation coefficient (r): 0.55952261014232 : http://www.alcula.com/calculators/statistics/correlation-coefficient/

    Sample size: 30
    Mean x (x̄): 88.1
    Mean y (ȳ): 0.40061666666667
    Intercept (a): -0.19408755573084
    Slope (b): 0.006750331695772
    Regression line equation: y=0.006750331695772x-0.19408755573084 : http://www.alcula.com/calculators/statistics/linear-regression/

    Changed countries highlighted:

    IQ,Met% Country
    104,0.609 Denmark
    102,0.512 Ireland
    100,0.53 UK
    99,0.572 Estonia
    99,0.248 Korea
    97,0.529 Finland
    97,0.47 France
    97,0.304 Japan
    96,0.533 Hungary
    96,0.46 Italy
    96,0.43 Spain
    95,0.478 Russia
    91,0.25 Mongolia
    89,0.5 Iran
    87,0.45 Turkey
    86,0.416 Lebanon
    86,0.416 Syria
    86,0.45 Mexico
    85,0.221 Thailand
    85,0.57 Palestine
    84,0.243 China
    83,0.3 Vietnam
    79,0.232 India
    79,0.465 Pakistan
    79,0.2475 Ghana
    77,0.345 Kenya
    75,0.332 Nigeria
    74,0.297 Cambodia
    72,0.299 Tanzania
    68,0.33 Senegal
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  11. FKA Max says: • Website
    @FKA Max
    Thanks again, res.

    These continent level numbers (particularly East Asia) make me skeptical about a strong Met% IQ connection.
     
    I understand your skepticism.

    This does not surprise me. Although Afrosapiens’s technique is interesting, using it for between country IQ comparisons of countries with widely differing levels of development has an obvious issue with the EA-development correlation interfering with the EA-IQ correlation.
     
    Yes, that is correct. His Syria IQ of 71 is probably due to current the war in Syria, for example. Mexico is also affected by drug wars, etc. The low Pakistani results could be due to high cousin marriage rates, which depress intelligence, high fertility rates and resultant resource scarcity/depletion, religious instead of scientific/academic teaching/study, etc.

    I think the Met% IQ correlation will become clearer and more pronounced the more peaceful, less religious/more secular and the more per-capita prosperous the world and/or a population/group becomes.

    The populations with higher Met% will likely experience "Super-Flynn Effects", when conditions are favorable for them:

    Race/IQ: Super-Flynn Effects in Germans, Jews, and Hispanics

    The central argument of my piece had been that although GDP and IQ were highly correlated, the direction of causality might well be from the former to the latter, and this attracted much derision.
     
    - http://www.unz.com/runz/raceiq-super-flynn-effects-in-germans-jews-and-hispanics/

    Put another way, the populations/groups with lower Met% might have a lower IQ ceiling and potential, and only experience "Semi-Flynn Effects" even under favorable conditions.

    These are some of the reasons why I am skeptical of test results coming out of East Asia, in particular China:


    The National Center for Fair and Open Testing, a nonprofit known as FairTest, which advocates against the misuse and abuse of standardized tests, said Chinese test prep companies have reported cancellations in Taiwan, Macao, Hong Kong, Japan, Singapore and Shanghai.
     
    - https://www.washingtonpost.com/news/answer-sheet/wp/2017/09/07/security-breach-forces-act-test-cancellation-in-asia-and-weather-causes-some-in-u-s/


    But as with just about everything concerning modern China, the results should also be viewed with some distance and possible skepticism. The 5000+ students who were tested in China's biggest and most modern city may or may not be indicative of broader progress throughout the country (as the NYT story points out). Anyone who has had experience with schools and testing in China will want to know more about how these tests were administered, supervised, and scored.
     
    - https://www.theatlantic.com/national/archive/2010/12/on-those-stunning-shanghai-test-scores/67654/

    When I slightly alter/manipulate the sample by taking 85,0.135 Micronesia and 72,0.44 Papa New Guinea out of the sample, changing Syria from IQ 71 to Lebanon’s IQ 86 and Pakistan’s IQ 70 to India’s IQ 79, and Mexico from Met% 0.6 to Met% 0.43 like Spain, I get a correlation coefficient of:

    Sample size: 30

    Correlation coefficient (r): 0.55952261014232 : http://www.alcula.com/calculators/statistics/correlation-coefficient/

    Sample size: 30
    Mean x (x̄): 88.1
    Mean y (ȳ): 0.40061666666667
    Intercept (a): -0.19408755573084
    Slope (b): 0.006750331695772
    Regression line equation: y=0.006750331695772x-0.19408755573084 : http://www.alcula.com/calculators/statistics/linear-regression/

    Changed countries highlighted:

    [MORE]

    IQ,Met% Country
    104,0.609 Denmark
    102,0.512 Ireland
    100,0.53 UK
    99,0.572 Estonia
    99,0.248 Korea
    97,0.529 Finland
    97,0.47 France
    97,0.304 Japan
    96,0.533 Hungary
    96,0.46 Italy
    96,0.43 Spain
    95,0.478 Russia
    91,0.25 Mongolia
    89,0.5 Iran
    87,0.45 Turkey
    86,0.416 Lebanon
    86,0.416 Syria
    86,0.45 Mexico
    85,0.221 Thailand
    85,0.57 Palestine
    84,0.243 China
    83,0.3 Vietnam
    79,0.232 India
    79,0.465 Pakistan
    79,0.2475 Ghana
    77,0.345 Kenya
    75,0.332 Nigeria
    74,0.297 Cambodia
    72,0.299 Tanzania
    68,0.33 Senegal

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  12. Davide Piffer tells me that the correlation between group average IQs and his polygenic prediction is now 0.9

    Read More
    • Replies: @res
    Thanks! Do you know which populations he was using?

    To add some context: https://www.unz.com/jthompson/genetics-of-racial-differences-in-intelligence-updated/
    https://topseudoscience.wordpress.com/2017/06/02/new-genes-same-results-group-level-genotypic-intelligence-for-26-and-52-populations/

    If I understand correctly the 26 population group is 1000 Genomes: http://www.internationalgenome.org/faq/which-populations-are-part-your-study/

    The 52 population group in the link above is from ALFRED (right?): https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3245092/
    but the more recent RPubs documents refer to CEPH-HGDP (included in ALFRED?): http://www.cephb.fr/en/hgdp_panel.php
    Are those two references the same data by different names?

    I think the consistently high correlations of IQ and Piffer's successive PGS provide a good validation of his hypothesis that the smaller SNP subsets capture selection pressure.

    The one PGS which seems like an exception is the relatively low correlations seen in http://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."

    To preserve context, the rest of that paragraph was: "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)"
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  13. Factorize says:

    Wow!

    How well does his polygenic predict individual level IQs? I am not sure whether there are any public IQs with phenotypes, though this would be a helpful resource.

    Might you ask him to run his program to predict the IQs of a few individuals from the 1000 Genomes Project? Perhaps HG00096, HG00097, HG00099, HG00100, HG00101? I would love to have a calibration sample. It would also be of interest to have the polygenic scores for these individuals by phased chromosome. How much variance is there between homologous chromosomes? Looking at the likely points of recombination along chromosomes and determining polygenic scores for different scenarios might be of even greater interest.

    Read More
    • Replies: @res
    It's not what you are asking for, but the CEU and Yoruba plots in https://www.unz.com/jthompson/the-dna-of-genius-n2/ are somewhat in that vein.

    Was the compressed sensing height predictor ever made public? It would be interesting to see how that works in this methodology. It would also provide a good check for the SNP subset/full predictor correlation idea.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  14. res says:
    @James Thompson
    Davide Piffer tells me that the correlation between group average IQs and his polygenic prediction is now 0.9

    Thanks! Do you know which populations he was using?

    To add some context: https://www.unz.com/jthompson/genetics-of-racial-differences-in-intelligence-updated/

    https://topseudoscience.wordpress.com/2017/06/02/new-genes-same-results-group-level-genotypic-intelligence-for-26-and-52-populations/

    If I understand correctly the 26 population group is 1000 Genomes: http://www.internationalgenome.org/faq/which-populations-are-part-your-study/

    The 52 population group in the link above is from ALFRED (right?): https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3245092/
    but the more recent RPubs documents refer to CEPH-HGDP (included in ALFRED?): http://www.cephb.fr/en/hgdp_panel.php
    Are those two references the same data by different names?

    I think the consistently high correlations of IQ and Piffer’s successive PGS provide a good validation of his hypothesis that the smaller SNP subsets capture selection pressure.

    The one PGS which seems like an exception is the relatively low correlations seen in http://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.”

    To preserve context, the rest of that paragraph was: “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)”

    Read More
    • Replies: @phil
    Good question! The 1000 GENOMES data for East Asia are biased toward urban elites in Beijing, Tokyo, and Saigon. If the ALFRED data are used, how does one aggregate all the Chinese ethnicities to come up with a composite allele score for China? The populations sampled for ALFRED are not the same as the groups that were sampled for the IQ results reported by Lynn.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  15. res says:
    @Factorize
    Wow!

    How well does his polygenic predict individual level IQs? I am not sure whether there are any public IQs with phenotypes, though this would be a helpful resource.

    Might you ask him to run his program to predict the IQs of a few individuals from the 1000 Genomes Project? Perhaps HG00096, HG00097, HG00099, HG00100, HG00101? I would love to have a calibration sample. It would also be of interest to have the polygenic scores for these individuals by phased chromosome. How much variance is there between homologous chromosomes? Looking at the likely points of recombination along chromosomes and determining polygenic scores for different scenarios might be of even greater interest.

    It’s not what you are asking for, but the CEU and Yoruba plots in https://www.unz.com/jthompson/the-dna-of-genius-n2/ are somewhat in that vein.

    Was the compressed sensing height predictor ever made public? It would be interesting to see how that works in this methodology. It would also provide a good check for the SNP subset/full predictor correlation idea.

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

    res, I am not sure what to make of this result (i.e., the 0.9 polygenic result). It is a fairly startling. utu really made a run at it last time round. I am going to fence sit on this one. It does make sense from the perspective of polygenic adaption. When people are choosing mates they will choose the cognitive phenotype that is observable and it does not matter if the variants involved have super tiny effect sizes. All the variants act as one package. I am just not sure about the technical questions involved. Are we missing something? It would be amazing if something like this could be applied at the individual level. That would really move things into gear! I think the current polygenic scoring might only be explaining perhaps up to 10% of IQ. Is it really possible that this new method could do much much better than that? Would be very startling if it could! Though I guess this is more about looking at the scale of population differences.

    Did he say that the SNPs he was using were those with maximal discriminatory power? I think utu made an argument that any set of SNPs would do. Nevertheless this is super exciting and I certainly wish some experts would provide clarification on this methodology.

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter Display All Comments
  17. I don’t understand why Africa [all africans*] was higher than America [amerindians].

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter Display All Comments
  18. Factorize says:

    res, one thing that is confusing me is that the polygenic result proposed is a falsifiable assertion. I would like to go at this more from a gedanken point of view, though a brute force empiricism would also give an unequivocal answer.

    This reduces to: Does the polygenic approach proposed replicate in independent preferably by independent researchers? I think that this is an important question to have clarified given the not entirely inconsequential implications if suggested results were indeed found to be accurate.

    Read More
    • Replies: @res

    Does the polygenic approach proposed replicate in independent preferably by independent researchers?
     
    Piffer has computed PGS from multiple studies (including different researchers) and they seem to replicate quite well (though I am curious about the apparent exception--lower correlation--noted at the end of comment 14). Also see his Monte Carlo simulation technique discussed in other posts which quantifies how unlikely it is that a random set of SNPs would give a similar result.

    His height work provides another form of replication: https://f1000research.com/articles/4-15/v3

    P.S. While looking at something else today I ran across a file which has forward/reverse (relative to dbSnp) strand data for the SNP chip used in the UKBB (1.1GB uncompressed csv, ouch!): www.ukbiobank.ac.uk/scientists-3/uk-biobank-axiom-array/
    It also has reference/alternate allele information (not sure if that is what the researchers typically base that on). That might be helpful for your PGS work.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  19. res says:
    @Factorize
    res, one thing that is confusing me is that the polygenic result proposed is a falsifiable assertion. I would like to go at this more from a gedanken point of view, though a brute force empiricism would also give an unequivocal answer.

    This reduces to: Does the polygenic approach proposed replicate in independent preferably by independent researchers? I think that this is an important question to have clarified given the not entirely inconsequential implications if suggested results were indeed found to be accurate.

    Does the polygenic approach proposed replicate in independent preferably by independent researchers?

    Piffer has computed PGS from multiple studies (including different researchers) and they seem to replicate quite well (though I am curious about the apparent exception–lower correlation–noted at the end of comment 14). Also see his Monte Carlo simulation technique discussed in other posts which quantifies how unlikely it is that a random set of SNPs would give a similar result.

    His height work provides another form of replication: https://f1000research.com/articles/4-15/v3

    P.S. While looking at something else today I ran across a file which has forward/reverse (relative to dbSnp) strand data for the SNP chip used in the UKBB (1.1GB uncompressed csv, ouch!): http://www.ukbiobank.ac.uk/scientists-3/uk-biobank-axiom-array/
    It also has reference/alternate allele information (not sure if that is what the researchers typically base that on). That might be helpful for your PGS work.

    Read More
    • Replies: @Steve Sailer
    Does Piffer's approach work on height?

    I like the idea of working out the kinks using the slightly less politically fraught topic of height.

    I'd like to see a test of nurture where we compare actual phenotypic height in various places to polygenic scores for height to see which polities come closer to fulfilling their people's genetic potential for height. It would seem like a good report card.

    It seems like we are pretty close to being able to do this for height.

    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  20. Davide Piffer said he would try to comment shortly, though he is tied up with other work at the moment.

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter Display All Comments
  21. FKA Max says: • Website
    @res
    Thanks. I downloaded the data for rs4680 from Piffer's source at http://spsmart.cesga.es/ceph.php?dataSet=ceph_stanford
    which allows me to look at COMT Met% for the exact populations Piffer used. Data below.

    I'll probably try to map the Lynn and Afrosapiens IQ estimates into Piffer's populations as I described in my earlier comment (unless you do it first). I am a bit surprised he did not do the Lynn comparison himself in the correlation page I linked.


    It actually weakened the correlation from r = 0.579 Lynn to r = 0.415 Afrosapiens.
     
    This does not surprise me. Although Afrosapiens's technique is interesting, using it for between country IQ comparisons of countries with widely differing levels of development has an obvious issue with the EA-development correlation interfering with the EA-IQ correlation. I think it is much better for within country comparisons (as in a typical EA GWAS) or countries with similar levels of development. Which makes it ironic that Afrosapiens touts his method in the worldwide IQ comparison context but was strangely silent when I brought it up to argue for an EA-IQ link in Nigeria (in a Chanda Chisala thread). That exchange helped make clear just how selective and motivated his reasoning is.

    P.S. Some links to the Nigeria EA-IQ conversation: http://www.unz.com/article/my-last-word-on-the-scrabble-and-iq-debate-2/#comment-2016734
    http://www.unz.com/article/my-last-word-on-the-scrabble-and-iq-debate-2/#comment-2012772
    I still can't believe Chanda tried to argue for a zero or even negative correlation between EA and IQ in Nigeria... SMH

    P.P.S. COMT Met% (rs4680 A allele) for Piffer populations below.



    Continent level:

    population N freq_A
    Population Set 1 944 0.389
    AFRICA 102 0.275
    AMERICA 64 0.305
    EUROPE 158 0.494
    MIDDLE EAST 163 0.463
    CENTRAL-SOUTH ASIA 200 0.477
    OCEANIA 28 0.393
    EAST ASIA 229 0.26

    These continent level numbers (particularly East Asia) make me skeptical about a strong Met% IQ connection.

    Subpopulation level:

    population N freq_A
    Population Set 1 944 0.389
    C. African Republic - Biaka Pygmy 22 0.068
    D. R. of Congo - Mbuti Pygmy 13 0.231
    Kenya - Bantu 11 0.364
    Namibia - San 5 0.1
    Nigeria - Yoruba 21 0.405
    Senegal - Mandenka 22 0.364
    South Africa - Bantu 8 0.313
    Brazil - Karitiana 14 0.036
    Brazil - Surui 8 0.313
    Colombia - Piapoco and Curripaco 7 0
    Mexico - Maya 21 0.571
    Mexico - Pima 14 0.321
    France - Basque 24 0.583
    France - French 28 0.482
    Italy - Sardinian 28 0.357
    Italy - Tuscan 8 0.5
    Italy - from Bergamo 13 0.423
    Orkney Islands - Orcadian 15 0.6
    Russia (Caucasus) - Adygei 17 0.412
    Russia - Russian 25 0.6
    Algeria (Mzab) - Mozabite 29 0.431
    Israel (Carmel) - Druze 42 0.464
    Israel (Central) - Palestinian 46 0.533
    Israel (Negev) - Bedouin 46 0.413
    China - Uygur 10 0.45
    Pakistan - Balochi 24 0.542
    Pakistan - Brahui 25 0.42
    Pakistan - Burusho 25 0.48
    Pakistan - Hazara 22 0.523
    Pakistan - Kalash 23 0.565
    Pakistan - Makrani 25 0.46
    Pakistan - Pathan 22 0.432
    Pakistan - Sindhi 24 0.417
    Bougainville - NAN Melanesian 11 0.318
    New Guinea - Papuan 17 0.441
    Cambodia - Cambodian 10 0.35
    China - Dai 10 0.2
    China - Daur 9 0.278
    China - Han 44 0.25
    China - Hezhen 9 0.333
    China - Lahu 8 0.313
    China - Miaozu 10 0.15
    China - Mongola 10 0.25
    China - Naxi 8 0.125
    China - Oroqen 9 0.222
    China - She 10 0.3
    China - Tu 10 0.3
    China - Tujia 10 0.2
    China - Xibo 9 0.167
    China - Yizu 10 0.25
    Japan - Japanese 28 0.304
    Siberia - Yakut 25 0.3

    These continent level numbers (particularly East Asia) make me skeptical about a strong Met% IQ connection.

    res,

    I found another study that could shed some light on what we were discussing, and actually, I believe, proves both of us right to some extent:

    Association study between COMT 158Met and creativity scores in bipolar disorder and healthy controls

    There are many difficulties inherent to systematic studies of creativity, particularly methodological problems concerning the reliability and validity of creativity measures, and disagreements over the definition of creativity. The BWAS is not the only measure of creativity and other measures should be explored before drawing more definitive conclusions. In the present study, it was decided to examine the correlation between scores on a widely used scale for measuring creativity and the presence of functional polymorphism of COMT (rs4680), which likely influences PFC cognition, in a homogeneous sample of university students. Our results are also consistent with those reported in the literature investigating the role of DA and COMT in PFC function and cognition. However, no influence of COMT on IQ was evident, and BWAS and IQ scores were unrelated, further suggesting some degree of specificity in the association of COMT with creativity.

    This study is the first to report findings that suggest the effects of COMT gene polymorphism may not be limited to isolated basal cognitive abilities, but could partially account for greater cognitive abilities related to creativity in healthy controls.

    http://www.scielo.br/scielo.php?script=sci_arttext&pid=S0101-60832014000200029

    Another Half Brick of Creativity

    The paper is slightly unusual, in that it seeks to model creative processes using IQ120 as a cut-off, and so runs contrary to the general findings of the Lubinski and Benbow work that there is no cut-off point, and that the brighter you are the more creative you are in real life.

    http://www.unz.com/jthompson/another-half-brick-of-creativity/#p_1_9

    Heave Half a Brick at Creativity

    Source: https://www.unz.com/jthompson/heave-half-brick-at-creativity/

    This raises several questions for me, specifically whether creativity is a better measure of intelligence/brightness than IQ is?

    Population COMT Met frequencies might be the best indicator we have to determine creativity, and lower COMT Met frequencies in East Asian populations might also explain the “low-creativity” stereotype often associated with East Asians.

    ‘Why Do Chinese Lack Creativity?’

    It’s not for lack of trying. The Chinese government now pours billions of dollars annually into research and development — by one estimate, its research and development budget may surpass U.S. spending by 2019 — and Chinese President Xi Jinping has emphasized innovation in his speeches. For the past four years, China has filed more patent applications than any other country, although state news agency Xinhua has described the quality of those patents as “poor.”http://foreignpolicy.com/2015/06/23/china-innovation-creativity-research-patents/

    I discussed something similar a while back with Afrosapiens:

    Following is my position on IQ tests. Maybe instead of “IQ test”, it should more accurately be called a “ Formal Education Potential, Quality and Attainment test” or something like that. I still feel the tests are useful and can offer some interesting insights, but they surely are and should not be the be-all and end-all tool of how we organize and structure our nations and societies, IMHO:

    I think what is important to reiterate is that IQ tests seem to measure and predict certain things very accurately, e.g., better cognitive performance/functioning under pressure/stress, educational attainment, income, possibly testosterone and dopamine levels, etc., but they do ironically/paradoxically only seem to test “intelligence” to a limited extent, at least that is my best, current understanding and interpretation of the data I have researched thus far.

    http://www.unz.com/jthompson/the-worlds-iq-86/#comment-2072872

    To conclude:

    What are Piffer et al.’s SNPs actually measuring?

    Read More
    • Replies: @RaceRealist88
    "What are Piffer et al.’s SNPs actually measuring?"

    Population stratification. Due to genetic drift etc it biases PRS scores for other non-European populations. It's just capturing population stratification.

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

    This is great! It really is who you know! I think I might have hit something with the empirical approach. The last round of this argument was a real grudge match. Why be like that? If those who want to disprove the polygenic approach suggested, it should be easy! Find a counter example. The code is out there; there should be datasets and phenotypes. If there’s a calculation error somewhere this should be cleared up in no time. However, as has been noted previously, if there is no dog bark soon, then after eliminating the impossible, all that will remain is the truth: the approach has achieved what it has claimed. That would be remarkable!

    Read More
    • Replies: @James Thompson
    Agreed. Expose any errors, test on new data sets. If it holds together, well then, until something else turns up, we can consider it to be right.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  23. @Factorize
    This is great! It really is who you know! I think I might have hit something with the empirical approach. The last round of this argument was a real grudge match. Why be like that? If those who want to disprove the polygenic approach suggested, it should be easy! Find a counter example. The code is out there; there should be datasets and phenotypes. If there's a calculation error somewhere this should be cleared up in no time. However, as has been noted previously, if there is no dog bark soon, then after eliminating the impossible, all that will remain is the truth: the approach has achieved what it has claimed. That would be remarkable!

    Agreed. Expose any errors, test on new data sets. If it holds together, well then, until something else turns up, we can consider it to be right.

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  24. dux.ie says:
    @res
    Sounds like something worth testing. I think the best way to do this is to incorporate the additional data into the analysis in http://rpubs.com/Daxide/377423
    I have that running locally so all I need is the appropriate data.

    The easiest way to do that would be to modify the HGDP_PGS .csv available at https://osf.io/uays8/
    Note the specific populations needed in the first column.

    Would it be possible for you to create a file adding the fields you want to that? I assume you would be looking at Lynn IQ, Afro IQ, and % Met?

    I can create the data file, but there are enough population mapping issues that I would prefer the person interested in the results make those assumptions.

    P.S. If Davide would like to look at this himself IMHO that would be even better.

    P.P.S. Do you have any thoughts why rs4680 does not show up in the results for the latest EA study? https://www.snpedia.com/index.php/Rs4680
    Though that page does show an IQ related connection: https://www.ncbi.nlm.nih.gov/pubmed/24853458?dopt=Abstract
    There is one SNP in the EA study that is in fairly high LD (D' = 0.8, r^2 =0.28 in CEU) with rs4680: rs2240715, but the p value is 0.7 with a small beta (I did not check exhaustively, just a quick visual look at chr 22 hits). Actually, rs165655 is also similar.

    Re: P.P.S. Do you have any thoughts why rs4680 does not show up in the results for the latest EA study? https://www.snpedia.com/index.php/Rs4680

    I have been saying this for a few years now. If you read the snpedia article closely, for warrior gene it talks of “higher pain threshold, better stress resiliency (in work places and group competitions), albeit with a modest reduction in executive cognition performance under most conditions (in home and individual undertakings)”, the warrior gene is more like a ‘grit gene’ than a ‘IQ gene’. The effects of rs4680 on IQ seems to be secondary. The performance of person with high grit without accompany high IQ will only be modest. The effects of ‘performance anxiety’ under group competitions can be seen in the OECD PISA data.

    The PISA project have data on the percentage of students who wanted to be the best, the higher that percentage the higher the competitive stress or anxiety, the results is shown in the chart,

    http://tinypic.com/view.php?pic=2v8fztg&s=9

    It can be seen that on average overall competitive stress reduces performance. However in residual analysis there is a dense cluster of outliers indicating that there are finer grain relationship which opposes the overall downward trend, i.e. the effects of stress which might be influenced by rs4680 is non-linear and exibiting a pitch-fork shape. The finer grain trend can be determined objectively and simply by dividing the data into 4 equal size quadrants and the data in the first quadrant gives the opposing trend.It is interesting that the countries in the first quadrant are the EastAsians, the Scandinavian countries from the descendents of the Vikings, the recently frontier countries or home countries of the descendents of the Anglo-Saxon/Celtic/Viking. The odd country was Japan whose students were not competitive and that had been identified in various studies as due to the national angst from losing WW2 that there could be epigenetic changes to the genomes the effects of which might last a few generations.

    The chart of PISA scores against the percentage of population percentage of rs4680 variant,

    http://tinypic.com/view.php?pic=nv9ph2&s=9

    If there is no overall statistical trends but simple clustering of the data space into two equal sets can produce two significant opposing trends, the hypothesis that there is no statistically significant result should be rejected. Note the rs4680 data for some countries in the lower fork with low PISA3 scores and most probably high FreqG are not available and the results showed a V shape. High performance can be achieved in both extreme high and low FreqG (or FreqA), thus it is primarily not an IQ gene.

    GWAS with the one track mind of only considering linear trends, the non-linear trends can easily be omitted.

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  25. phil says:
    @res
    Thanks! Do you know which populations he was using?

    To add some context: https://www.unz.com/jthompson/genetics-of-racial-differences-in-intelligence-updated/
    https://topseudoscience.wordpress.com/2017/06/02/new-genes-same-results-group-level-genotypic-intelligence-for-26-and-52-populations/

    If I understand correctly the 26 population group is 1000 Genomes: http://www.internationalgenome.org/faq/which-populations-are-part-your-study/

    The 52 population group in the link above is from ALFRED (right?): https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3245092/
    but the more recent RPubs documents refer to CEPH-HGDP (included in ALFRED?): http://www.cephb.fr/en/hgdp_panel.php
    Are those two references the same data by different names?

    I think the consistently high correlations of IQ and Piffer's successive PGS provide a good validation of his hypothesis that the smaller SNP subsets capture selection pressure.

    The one PGS which seems like an exception is the relatively low correlations seen in http://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."

    To preserve context, the rest of that paragraph was: "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)"

    Good question! The 1000 GENOMES data for East Asia are biased toward urban elites in Beijing, Tokyo, and Saigon. If the ALFRED data are used, how does one aggregate all the Chinese ethnicities to come up with a composite allele score for China? The populations sampled for ALFRED are not the same as the groups that were sampled for the IQ results reported by Lynn.

    Read More
    • Replies: @res
    For the Chinese ethnicities I think a population weighted average makes sense. Given that the Han are >90% of the population, that pretty much turns into "look at the Han number." https://www.travelchinaguide.com/intro/nationality/
    Do you have a sense of what % of Chinese population is covered by groups tested by both ALFRED and Lynn?

    That does beg the question of how much intra-Han variation there is. This was the best paper I saw about that: https://www.biorxiv.org/content/early/2017/07/13/162982
    But I am skeptical of their median 1.7x coverage (what do you think?). Figure 2 has the PCA population structure, but the % variance explained of PC1 and PC2 is tiny (0.146% and 0.032%) compared to continental races. I am not sure how it compares to e.g. North-South Italian differences.

    Mapping IQ data into the various population groups seems to me like the most questionable (vulnerable to both measurement/mapping error and bias) aspect of his research. I get the sense Piffer is making an honest effort here, but I would feel better if there was more transparent discussion of how the IQ numbers and PGS results are assigned for the different populations. Ethnic groups and countries can be quite different. Perhaps this discussion is embedded in his various earlier papers, but I think it would be good to make it more explicit for each set of data being used.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  26. res says:
    @phil
    Good question! The 1000 GENOMES data for East Asia are biased toward urban elites in Beijing, Tokyo, and Saigon. If the ALFRED data are used, how does one aggregate all the Chinese ethnicities to come up with a composite allele score for China? The populations sampled for ALFRED are not the same as the groups that were sampled for the IQ results reported by Lynn.

    For the Chinese ethnicities I think a population weighted average makes sense. Given that the Han are >90% of the population, that pretty much turns into “look at the Han number.” https://www.travelchinaguide.com/intro/nationality/
    Do you have a sense of what % of Chinese population is covered by groups tested by both ALFRED and Lynn?

    That does beg the question of how much intra-Han variation there is. This was the best paper I saw about that: https://www.biorxiv.org/content/early/2017/07/13/162982
    But I am skeptical of their median 1.7x coverage (what do you think?). Figure 2 has the PCA population structure, but the % variance explained of PC1 and PC2 is tiny (0.146% and 0.032%) compared to continental races. I am not sure how it compares to e.g. North-South Italian differences.

    Mapping IQ data into the various population groups seems to me like the most questionable (vulnerable to both measurement/mapping error and bias) aspect of his research. I get the sense Piffer is making an honest effort here, but I would feel better if there was more transparent discussion of how the IQ numbers and PGS results are assigned for the different populations. Ethnic groups and countries can be quite different. Perhaps this discussion is embedded in his various earlier papers, but I think it would be good to make it more explicit for each set of data being used.

    Read More
    • Replies: @phil
    Whether it's neglect of ethnic minorities, or rural Han people, the samples used by Lynn and Vanhanen for China are biased. The PISA results by province indicate an average IQ for China that is still probably above 100, but not really close to 105. On the other hand, there is still a positive Flynn effect going on in China, despite some dysgenic fertility.

    Japan represents a different challenge. The cognitive allele frequencies are not really that impressive compared to Utah whites, and yet they may be biased toward urban areas. On the other hand, there is probably a substantial difference in average IQ between Japanese and Utah whites.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  27. Factorize says:

    res, this is very exciting! It is surprising to me that there seems to be such a silence in the scientific community about this. It is not so much that people are arguing with the results; they are simply ignoring them. It is somewhat miraculous that this approach would produce the results that it has; I am greatly looking forward to seeing how this is resolved.

    Yes, in terms of the forward/reverse strand I was thrown for a loop by that. I think it was noted earlier on this blog. I thought linking the effect allele with the genotype allele made a great deal of sense. Are they really saying that the reported genotype might not be the actual genotype? I found this very confusing.

    I am very much looking forward to having a look at the polygenic scores by chromosome. Any guess on what typical strand differences in chromosomal PGS scores might exist? As a rough guess, I suggest perhaps 3-7 points as the largest such differences. I would love to have a calibration sample!

    Read More
    • Replies: @RaceRealist88
    "res, this is very exciting! It is surprising to me that there seems to be such a silence in the scientific community about this"

    Because a host of problems exist. Are you aware of the schizophrenia PRS paper regarding ancestry?
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  28. phil says:
    @res
    For the Chinese ethnicities I think a population weighted average makes sense. Given that the Han are >90% of the population, that pretty much turns into "look at the Han number." https://www.travelchinaguide.com/intro/nationality/
    Do you have a sense of what % of Chinese population is covered by groups tested by both ALFRED and Lynn?

    That does beg the question of how much intra-Han variation there is. This was the best paper I saw about that: https://www.biorxiv.org/content/early/2017/07/13/162982
    But I am skeptical of their median 1.7x coverage (what do you think?). Figure 2 has the PCA population structure, but the % variance explained of PC1 and PC2 is tiny (0.146% and 0.032%) compared to continental races. I am not sure how it compares to e.g. North-South Italian differences.

    Mapping IQ data into the various population groups seems to me like the most questionable (vulnerable to both measurement/mapping error and bias) aspect of his research. I get the sense Piffer is making an honest effort here, but I would feel better if there was more transparent discussion of how the IQ numbers and PGS results are assigned for the different populations. Ethnic groups and countries can be quite different. Perhaps this discussion is embedded in his various earlier papers, but I think it would be good to make it more explicit for each set of data being used.

    Whether it’s neglect of ethnic minorities, or rural Han people, the samples used by Lynn and Vanhanen for China are biased. The PISA results by province indicate an average IQ for China that is still probably above 100, but not really close to 105. On the other hand, there is still a positive Flynn effect going on in China, despite some dysgenic fertility.

    Japan represents a different challenge. The cognitive allele frequencies are not really that impressive compared to Utah whites, and yet they may be biased toward urban areas. On the other hand, there is probably a substantial difference in average IQ between Japanese and Utah whites.

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

    res, I finally thought about the strand flip issue carefully and it should not be very difficult to program around it after all. If the effect allele does not show up in the genotypes, then I could simply take the complements of the genotypes. I am just not sure whether there might be some situations that could be ambiguous. What if the effect allele were T and the genotypes were A and T. Sometimes the frequencies that they gave appear different from dbsnp, so resolving some problems might not be obvious. It is great that the software has been uploaded to calculate the PGS from the paper. It might be easiest if I were to fork the existing code to do what I want. I would need to find a way to organize the PGS as a total chromosome score and not a genome wide score.

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter Display All Comments
  30. @FKA Max


    These continent level numbers (particularly East Asia) make me skeptical about a strong Met% IQ connection.
     
    res,

    I found another study that could shed some light on what we were discussing, and actually, I believe, proves both of us right to some extent:

    Association study between COMT 158Met and creativity scores in bipolar disorder and healthy controls

    There are many difficulties inherent to systematic studies of creativity, particularly methodological problems concerning the reliability and validity of creativity measures, and disagreements over the definition of creativity. The BWAS is not the only measure of creativity and other measures should be explored before drawing more definitive conclusions. In the present study, it was decided to examine the correlation between scores on a widely used scale for measuring creativity and the presence of functional polymorphism of COMT (rs4680), which likely influences PFC cognition, in a homogeneous sample of university students. Our results are also consistent with those reported in the literature investigating the role of DA and COMT in PFC function and cognition. However, no influence of COMT on IQ was evident, and BWAS and IQ scores were unrelated, further suggesting some degree of specificity in the association of COMT with creativity.

    This study is the first to report findings that suggest the effects of COMT gene polymorphism may not be limited to isolated basal cognitive abilities, but could partially account for greater cognitive abilities related to creativity in healthy controls.
     
    - http://www.scielo.br/scielo.php?script=sci_arttext&pid=S0101-60832014000200029

    Another Half Brick of Creativity


    The paper is slightly unusual, in that it seeks to model creative processes using IQ120 as a cut-off, and so runs contrary to the general findings of the Lubinski and Benbow work that there is no cut-off point, and that the brighter you are the more creative you are in real life.
     
    - http://www.unz.com/jthompson/another-half-brick-of-creativity/#p_1_9

    Heave Half a Brick at Creativity

    http://lh6.ggpht.com/-6RIi9RaWz0I/VKWLiLxk8fI/AAAAAAAAAuA/LpL_OgIuaec/image_thumb%25255B7%25255D.png

    Source: https://www.unz.com/jthompson/heave-half-brick-at-creativity/

    This raises several questions for me, specifically whether creativity is a better measure of intelligence/brightness than IQ is?

    Population COMT Met frequencies might be the best indicator we have to determine creativity, and lower COMT Met frequencies in East Asian populations might also explain the "low-creativity" stereotype often associated with East Asians.

    ‘Why Do Chinese Lack Creativity?’

    It’s not for lack of trying. The Chinese government now pours billions of dollars annually into research and development — by one estimate, its research and development budget may surpass U.S. spending by 2019 — and Chinese President Xi Jinping has emphasized innovation in his speeches. For the past four years, China has filed more patent applications than any other country, although state news agency Xinhua has described the quality of those patents as “poor.” - http://foreignpolicy.com/2015/06/23/china-innovation-creativity-research-patents/

    I discussed something similar a while back with Afrosapiens:

    Following is my position on IQ tests. Maybe instead of “IQ test”, it should more accurately be called a “ Formal Education Potential, Quality and Attainment test” or something like that. I still feel the tests are useful and can offer some interesting insights, but they surely are and should not be the be-all and end-all tool of how we organize and structure our nations and societies, IMHO:

    I think what is important to reiterate is that IQ tests seem to measure and predict certain things very accurately, e.g., better cognitive performance/functioning under pressure/stress, educational attainment, income, possibly testosterone and dopamine levels, etc., but they do ironically/paradoxically only seem to test “intelligence” to a limited extent, at least that is my best, current understanding and interpretation of the data I have researched thus far.
     
    - http://www.unz.com/jthompson/the-worlds-iq-86/#comment-2072872

    To conclude:

    What are Piffer et al.'s SNPs actually measuring?

    “What are Piffer et al.’s SNPs actually measuring?”

    Population stratification. Due to genetic drift etc it biases PRS scores for other non-European populations. It’s just capturing population stratification.

    Read More
    • Replies: @res

    It’s just capturing population stratification.
     
    Take a look at the Racimo paper mentioned above. Do you think that method is sufficient to prove or disprove your contention?
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  31. @Factorize
    res, this is very exciting! It is surprising to me that there seems to be such a silence in the scientific community about this. It is not so much that people are arguing with the results; they are simply ignoring them. It is somewhat miraculous that this approach would produce the results that it has; I am greatly looking forward to seeing how this is resolved.

    Yes, in terms of the forward/reverse strand I was thrown for a loop by that. I think it was noted earlier on this blog. I thought linking the effect allele with the genotype allele made a great deal of sense. Are they really saying that the reported genotype might not be the actual genotype? I found this very confusing.

    I am very much looking forward to having a look at the polygenic scores by chromosome. Any guess on what typical strand differences in chromosomal PGS scores might exist? As a rough guess, I suggest perhaps 3-7 points as the largest such differences. I would love to have a calibration sample!

    “res, this is very exciting! It is surprising to me that there seems to be such a silence in the scientific community about this”

    Because a host of problems exist. Are you aware of the schizophrenia PRS paper regarding ancestry?

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

    Does the polygenic approach proposed replicate in independent preferably by independent researchers?
     
    Piffer has computed PGS from multiple studies (including different researchers) and they seem to replicate quite well (though I am curious about the apparent exception--lower correlation--noted at the end of comment 14). Also see his Monte Carlo simulation technique discussed in other posts which quantifies how unlikely it is that a random set of SNPs would give a similar result.

    His height work provides another form of replication: https://f1000research.com/articles/4-15/v3

    P.S. While looking at something else today I ran across a file which has forward/reverse (relative to dbSnp) strand data for the SNP chip used in the UKBB (1.1GB uncompressed csv, ouch!): www.ukbiobank.ac.uk/scientists-3/uk-biobank-axiom-array/
    It also has reference/alternate allele information (not sure if that is what the researchers typically base that on). That might be helpful for your PGS work.

    Does Piffer’s approach work on height?

    I like the idea of working out the kinks using the slightly less politically fraught topic of height.

    I’d like to see a test of nurture where we compare actual phenotypic height in various places to polygenic scores for height to see which polities come closer to fulfilling their people’s genetic potential for height. It would seem like a good report card.

    It seems like we are pretty close to being able to do this for height.

    Read More
    • Replies: @James Thompson
    Piffer has done a test in which he sees whether the polygenic score for intelligence predicts height, (I know, please stick with this) just to check whether his factor is picking up something general about race, rather than something which homes in on intelligence. It does not predict height at all.
    The guy who has "cracked" height is Steve Hsu, who can predict it in European population to within about an inch and a quarter.
    As you say, doing for all this for height is less fraught. Yes, there is a good environmental story to be told about boosting height. In the end, it reaches a plateau, but it is a better place to be anyway.
    , @res

    Does Piffer’s approach work on height?
     
    Piffer worked on this a few years ago. This paper quotes 0.79 and 0.83 correlations for a limited number of countries: https://f1000research.com/articles/4-15/v3
    See Table 1 and the Polygenic score section for details.

    This recent document calculates a height PGS, but I don't see it related to phenotypic height: http://rpubs.com/Daxide/377423

    I like the idea of working out the kinks using the slightly less politically fraught topic of height.
     
    I strongly agree. Not only that, but also more accurate, less controversial, and more easily available phenotypic data. Height would be a much better framework for evaluating and refining the methodology.

    I’d like to see a test of nurture where we compare actual phenotypic height in various places to polygenic scores for height to see which polities come closer to fulfilling their people’s genetic potential for height. It would seem like a good report card.
     
    I like this idea. North and South Korea would make a great test case.

    It seems like we are pretty close to being able to do this for height.
     
    One issue is the population groups covered by the current height work. The Hsu et al. compressed sensing works seems to have (roughly) "solved" (found all R^2 predicted by GCTA) the additive genetic variance for height within a fairly homogenous portion of the UK BioBank population. Extending that to other populations requires large sample studies. I am not holding my breath waiting for an African country to do something like the UKBB. The Chinese on the other hand...
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  33. How different is Piffer’s approach from Racimo’s study of educational attainment, height, and unibrow?

    Detecting Polygenic Adaptation in Admixture Graphs
    Fernando Racimo, Jeremy J. Berg and Joseph K. Pickrell
    GENETICS Early online January 18, 2018; https://doi.org/10.1534/genetics.117.300489

    http://www.genetics.org/content/early/2018/01/18/genetics.117.300489

    Read More
    • Replies: @James Thompson
    Need to gain entry before I can comment.
    , @res
    I am not an expert here so would like to hear other opinions, but I think the Racimo approach is very different from Piffer's.

    Piffer's approach is much more direct and has the virtue of being testable with the correlations he observes.

    Racimo's approach is much more complex and has the virtue of providing details of trait selection throughout population branching. I find it fascinating that we are able to infer this.

    It is worth noting that if Piffer's hypothesis about the initially detected SNPs being representative of overall selection pressure holds then I think Racimo's approach should be informative with a fairly small number of SNPs. Similarly, I think Racimo's work is supportive of Piffer's hypothesis and could be even more so if they did comparisons between different sizes of SNP sets.

    Figure 7 on page 19 is an interesting look at allele frequencies and effect size. Here is an earlier version:
    https://2.bp.blogspot.com/-VbM059x_Zf4/WTf9mVG4lHI/AAAAAAAASfE/W8jvgYQy0c4xkmhU-2gFQMByhobyIypZgCLcB/s400/Screen%2BShot%2B2017-06-07%2Bat%2B9.19.47%2BAM.png

    I think Figure S38 on page 64/87 is the money figure for the paper (not sure why they used Figure 8 in the body instead). S36 and S37 are similar. It shows selection on the branches of the world population tree for the four traits. The results seem intuitive to me except I am a bit surprised there is no sign of positive selection on the r-q population branch for EA. It is possible that the SNPs underlying that selection (if it exists) would need an EA GWAS including both Europeans and Africans to detect.

    This link is different, but gives an idea of what the four S38 panels look like (I think the S38 height version is better because it also shows negative selection for height in Asia):

    https://github.com/FerRacimo/PolyGraph/raw/master/HEIGHT_1KG_YRI_CEU_CHB_PEL_CLM.png

    Racimo et al. provide R code for height at https://github.com/FerRacimo/PolyGraph
    It would be interesting to try the following experiments:
    - Replace their height SNPs with the compressed sensing version.
    - Replace their height SNPs with recent EA/IQ results.

    I need to try playing with that. Though perhaps using it with many SNP results is a bad idea: "The trace from the MCMC run (slower to compute; may take hours or days, depending on the complexity of the graph and the number of trait-affecting SNPs)" They used 532 height SNPs: https://github.com/FerRacimo/PolyGraph/blob/master/GWAS_HEIGHT_1000genomes_allpops.txt

    Thinking about it some more, I think the experiment to try is running their analysis on Piffer's initial (small) IQ/EA SNP sets.

    P.S. Dr. Thompson, not sure what you meant in comment 35, but the second link Steve gave has full text of the preprint (all 107 pages!).
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  34. @Steve Sailer
    Does Piffer's approach work on height?

    I like the idea of working out the kinks using the slightly less politically fraught topic of height.

    I'd like to see a test of nurture where we compare actual phenotypic height in various places to polygenic scores for height to see which polities come closer to fulfilling their people's genetic potential for height. It would seem like a good report card.

    It seems like we are pretty close to being able to do this for height.

    Piffer has done a test in which he sees whether the polygenic score for intelligence predicts height, (I know, please stick with this) just to check whether his factor is picking up something general about race, rather than something which homes in on intelligence. It does not predict height at all.
    The guy who has “cracked” height is Steve Hsu, who can predict it in European population to within about an inch and a quarter.
    As you say, doing for all this for height is less fraught. Yes, there is a good environmental story to be told about boosting height. In the end, it reaches a plateau, but it is a better place to be anyway.

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  35. @Steve Sailer
    How different is Piffer's approach from Racimo's study of educational attainment, height, and unibrow?

    Detecting Polygenic Adaptation in Admixture Graphs
    Fernando Racimo, Jeremy J. Berg and Joseph K. Pickrell
    GENETICS Early online January 18, 2018; https://doi.org/10.1534/genetics.117.300489

    http://www.genetics.org/content/early/2018/01/18/genetics.117.300489

    Need to gain entry before I can comment.

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  36. res says:
    @Steve Sailer
    Does Piffer's approach work on height?

    I like the idea of working out the kinks using the slightly less politically fraught topic of height.

    I'd like to see a test of nurture where we compare actual phenotypic height in various places to polygenic scores for height to see which polities come closer to fulfilling their people's genetic potential for height. It would seem like a good report card.

    It seems like we are pretty close to being able to do this for height.

    Does Piffer’s approach work on height?

    Piffer worked on this a few years ago. This paper quotes 0.79 and 0.83 correlations for a limited number of countries: https://f1000research.com/articles/4-15/v3
    See Table 1 and the Polygenic score section for details.

    This recent document calculates a height PGS, but I don’t see it related to phenotypic height: http://rpubs.com/Daxide/377423

    I like the idea of working out the kinks using the slightly less politically fraught topic of height.

    I strongly agree. Not only that, but also more accurate, less controversial, and more easily available phenotypic data. Height would be a much better framework for evaluating and refining the methodology.

    I’d like to see a test of nurture where we compare actual phenotypic height in various places to polygenic scores for height to see which polities come closer to fulfilling their people’s genetic potential for height. It would seem like a good report card.

    I like this idea. North and South Korea would make a great test case.

    It seems like we are pretty close to being able to do this for height.

    One issue is the population groups covered by the current height work. The Hsu et al. compressed sensing works seems to have (roughly) “solved” (found all R^2 predicted by GCTA) the additive genetic variance for height within a fairly homogenous portion of the UK BioBank population. Extending that to other populations requires large sample studies. I am not holding my breath waiting for an African country to do something like the UKBB. The Chinese on the other hand…

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  37. res says:
    @Steve Sailer
    How different is Piffer's approach from Racimo's study of educational attainment, height, and unibrow?

    Detecting Polygenic Adaptation in Admixture Graphs
    Fernando Racimo, Jeremy J. Berg and Joseph K. Pickrell
    GENETICS Early online January 18, 2018; https://doi.org/10.1534/genetics.117.300489

    http://www.genetics.org/content/early/2018/01/18/genetics.117.300489

    I am not an expert here so would like to hear other opinions, but I think the Racimo approach is very different from Piffer’s.

    Piffer’s approach is much more direct and has the virtue of being testable with the correlations he observes.

    Racimo’s approach is much more complex and has the virtue of providing details of trait selection throughout population branching. I find it fascinating that we are able to infer this.

    It is worth noting that if Piffer’s hypothesis about the initially detected SNPs being representative of overall selection pressure holds then I think Racimo’s approach should be informative with a fairly small number of SNPs. Similarly, I think Racimo’s work is supportive of Piffer’s hypothesis and could be even more so if they did comparisons between different sizes of SNP sets.

    Figure 7 on page 19 is an interesting look at allele frequencies and effect size. Here is an earlier version:

    I think Figure S38 on page 64/87 is the money figure for the paper (not sure why they used Figure 8 in the body instead). S36 and S37 are similar. It shows selection on the branches of the world population tree for the four traits. The results seem intuitive to me except I am a bit surprised there is no sign of positive selection on the r-q population branch for EA. It is possible that the SNPs underlying that selection (if it exists) would need an EA GWAS including both Europeans and Africans to detect.

    This link is different, but gives an idea of what the four S38 panels look like (I think the S38 height version is better because it also shows negative selection for height in Asia):

    Racimo et al. provide R code for height at https://github.com/FerRacimo/PolyGraph
    It would be interesting to try the following experiments:
    - Replace their height SNPs with the compressed sensing version.
    - Replace their height SNPs with recent EA/IQ results.

    I need to try playing with that. Though perhaps using it with many SNP results is a bad idea: “The trace from the MCMC run (slower to compute; may take hours or days, depending on the complexity of the graph and the number of trait-affecting SNPs)” They used 532 height SNPs: https://github.com/FerRacimo/PolyGraph/blob/master/GWAS_HEIGHT_1000genomes_allpops.txt

    Thinking about it some more, I think the experiment to try is running their analysis on Piffer’s initial (small) IQ/EA SNP sets.

    P.S. Dr. Thompson, not sure what you meant in comment 35, but the second link Steve gave has full text of the preprint (all 107 pages!).

    Read More
    • Replies: @James Thompson
    Thanks for your contribution. I had clicked the first link only, hence my remark. I agree that Racimo is taking a complex but interesting approach to one specific problem: how do we really establish that a change in allele frequencies is due to selection and not to other more general processes.
    As to checking Piffer's work on other populations in more depth, I think there is African data being sat on somewhere, but probably without additional material even on years of education. Agree we are unlikely to get continent wide good quality African data for quite a while yet.
    Personally, if Hsu has cracked height, then IQ are next in line for a speedy solution. It will be interesting if he can ever get close to a 4 IQ error range, which is after all the test-retest discrepancy.
    , @res
    I downloaded their R code to work on this. I was able to partially replicate their R results (I cut down the MCMC steps by a factor of 500, which still took me about 10 minutes to run). The results from that directory are a very small subset of what was presented in the paper.

    One thing I was not able to figure out is their code loads the parallel package, but does not appear to support multicore CPUs in any way I could see. Sorting that out would be a big help for trying a full MCMC run.

    They only included the graph format of the last plot above so I was unable to look at the other representations.

    The major problem is that their analysis requires information about the "number of ancestral and derived alleles in each population that is a leaf in our graph." Since they supplied a neutral (wrt phenotype in question) input file I had hoped to be able to use that file as a source for other SNP data. Unfortunately, it looks like they trimmed that file (for all phenotypes they looked at, or ?) so it does not have any of Piffer's 9 IQ/EA SNPs and only has 78 of Lee's ~10k EA SNPs.

    So with the files they distributed it does not look like it is possible to evaluate any IQ/EA data. Since the files supplied were for height it may be possible to look at earlier height GWAS results to check how their SNP results compare to the 531 SNP version used in the R code. It is necessary to have the SNP effect sizes from the GWAS though.
    , @res
    I was looking through Racimo's paper again today and took a closer look at two very interesting supplemental graphics.

    Figure S56 on page 105 is a variant of the idea for validating Piffer's method I have discussed where one looks at successively more comprehensive sets of SNPs with a PGS. The figure gives genetic scores by 1000 Genomes population for the UKBB for p-values of 10^-9, 10^-8, and 10^-7. Visually we see good correspondence with some variation. It should be possible to match those with the phenotypic country values in a fashion similar to Piffer's and check how the correlations vary.

    Perhaps a better variation would be to look at mutually exclusive groups of SNPs (e.g. bucket by p-value) in a similar fashion. This would prevent the issue of the intercorrelations of variables like A, A+B, A+B+C.

    Figure S58 compares graphics built using two different selection measures. IIRC their git code uses the qb statistics to initialize their MCMC calculations of the alpha parameters, but I am not clear on the relative strengths and weaknesses of the estimates.
    The qb statistics show higher levels of selection signals than the alpha parameters (and in a way that matches intuition for S58 height). I am guessing they are being conservative by showing the alpha parameters (in most/all? of the other graphics), but the qb statistics seem more informative and are much easier/faster to calculate.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  38. res says:
    @RaceRealist88
    "What are Piffer et al.’s SNPs actually measuring?"

    Population stratification. Due to genetic drift etc it biases PRS scores for other non-European populations. It's just capturing population stratification.

    It’s just capturing population stratification.

    Take a look at the Racimo paper mentioned above. Do you think that method is sufficient to prove or disprove your contention?

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  39. @res
    I am not an expert here so would like to hear other opinions, but I think the Racimo approach is very different from Piffer's.

    Piffer's approach is much more direct and has the virtue of being testable with the correlations he observes.

    Racimo's approach is much more complex and has the virtue of providing details of trait selection throughout population branching. I find it fascinating that we are able to infer this.

    It is worth noting that if Piffer's hypothesis about the initially detected SNPs being representative of overall selection pressure holds then I think Racimo's approach should be informative with a fairly small number of SNPs. Similarly, I think Racimo's work is supportive of Piffer's hypothesis and could be even more so if they did comparisons between different sizes of SNP sets.

    Figure 7 on page 19 is an interesting look at allele frequencies and effect size. Here is an earlier version:
    https://2.bp.blogspot.com/-VbM059x_Zf4/WTf9mVG4lHI/AAAAAAAASfE/W8jvgYQy0c4xkmhU-2gFQMByhobyIypZgCLcB/s400/Screen%2BShot%2B2017-06-07%2Bat%2B9.19.47%2BAM.png

    I think Figure S38 on page 64/87 is the money figure for the paper (not sure why they used Figure 8 in the body instead). S36 and S37 are similar. It shows selection on the branches of the world population tree for the four traits. The results seem intuitive to me except I am a bit surprised there is no sign of positive selection on the r-q population branch for EA. It is possible that the SNPs underlying that selection (if it exists) would need an EA GWAS including both Europeans and Africans to detect.

    This link is different, but gives an idea of what the four S38 panels look like (I think the S38 height version is better because it also shows negative selection for height in Asia):

    https://github.com/FerRacimo/PolyGraph/raw/master/HEIGHT_1KG_YRI_CEU_CHB_PEL_CLM.png

    Racimo et al. provide R code for height at https://github.com/FerRacimo/PolyGraph
    It would be interesting to try the following experiments:
    - Replace their height SNPs with the compressed sensing version.
    - Replace their height SNPs with recent EA/IQ results.

    I need to try playing with that. Though perhaps using it with many SNP results is a bad idea: "The trace from the MCMC run (slower to compute; may take hours or days, depending on the complexity of the graph and the number of trait-affecting SNPs)" They used 532 height SNPs: https://github.com/FerRacimo/PolyGraph/blob/master/GWAS_HEIGHT_1000genomes_allpops.txt

    Thinking about it some more, I think the experiment to try is running their analysis on Piffer's initial (small) IQ/EA SNP sets.

    P.S. Dr. Thompson, not sure what you meant in comment 35, but the second link Steve gave has full text of the preprint (all 107 pages!).

    Thanks for your contribution. I had clicked the first link only, hence my remark. I agree that Racimo is taking a complex but interesting approach to one specific problem: how do we really establish that a change in allele frequencies is due to selection and not to other more general processes.
    As to checking Piffer’s work on other populations in more depth, I think there is African data being sat on somewhere, but probably without additional material even on years of education. Agree we are unlikely to get continent wide good quality African data for quite a while yet.
    Personally, if Hsu has cracked height, then IQ are next in line for a speedy solution. It will be interesting if he can ever get close to a 4 IQ error range, which is after all the test-retest discrepancy.

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  40. res says:
    @res
    I am not an expert here so would like to hear other opinions, but I think the Racimo approach is very different from Piffer's.

    Piffer's approach is much more direct and has the virtue of being testable with the correlations he observes.

    Racimo's approach is much more complex and has the virtue of providing details of trait selection throughout population branching. I find it fascinating that we are able to infer this.

    It is worth noting that if Piffer's hypothesis about the initially detected SNPs being representative of overall selection pressure holds then I think Racimo's approach should be informative with a fairly small number of SNPs. Similarly, I think Racimo's work is supportive of Piffer's hypothesis and could be even more so if they did comparisons between different sizes of SNP sets.

    Figure 7 on page 19 is an interesting look at allele frequencies and effect size. Here is an earlier version:
    https://2.bp.blogspot.com/-VbM059x_Zf4/WTf9mVG4lHI/AAAAAAAASfE/W8jvgYQy0c4xkmhU-2gFQMByhobyIypZgCLcB/s400/Screen%2BShot%2B2017-06-07%2Bat%2B9.19.47%2BAM.png

    I think Figure S38 on page 64/87 is the money figure for the paper (not sure why they used Figure 8 in the body instead). S36 and S37 are similar. It shows selection on the branches of the world population tree for the four traits. The results seem intuitive to me except I am a bit surprised there is no sign of positive selection on the r-q population branch for EA. It is possible that the SNPs underlying that selection (if it exists) would need an EA GWAS including both Europeans and Africans to detect.

    This link is different, but gives an idea of what the four S38 panels look like (I think the S38 height version is better because it also shows negative selection for height in Asia):

    https://github.com/FerRacimo/PolyGraph/raw/master/HEIGHT_1KG_YRI_CEU_CHB_PEL_CLM.png

    Racimo et al. provide R code for height at https://github.com/FerRacimo/PolyGraph
    It would be interesting to try the following experiments:
    - Replace their height SNPs with the compressed sensing version.
    - Replace their height SNPs with recent EA/IQ results.

    I need to try playing with that. Though perhaps using it with many SNP results is a bad idea: "The trace from the MCMC run (slower to compute; may take hours or days, depending on the complexity of the graph and the number of trait-affecting SNPs)" They used 532 height SNPs: https://github.com/FerRacimo/PolyGraph/blob/master/GWAS_HEIGHT_1000genomes_allpops.txt

    Thinking about it some more, I think the experiment to try is running their analysis on Piffer's initial (small) IQ/EA SNP sets.

    P.S. Dr. Thompson, not sure what you meant in comment 35, but the second link Steve gave has full text of the preprint (all 107 pages!).

    I downloaded their R code to work on this. I was able to partially replicate their R results (I cut down the MCMC steps by a factor of 500, which still took me about 10 minutes to run). The results from that directory are a very small subset of what was presented in the paper.

    One thing I was not able to figure out is their code loads the parallel package, but does not appear to support multicore CPUs in any way I could see. Sorting that out would be a big help for trying a full MCMC run.

    They only included the graph format of the last plot above so I was unable to look at the other representations.

    The major problem is that their analysis requires information about the “number of ancestral and derived alleles in each population that is a leaf in our graph.” Since they supplied a neutral (wrt phenotype in question) input file I had hoped to be able to use that file as a source for other SNP data. Unfortunately, it looks like they trimmed that file (for all phenotypes they looked at, or ?) so it does not have any of Piffer’s 9 IQ/EA SNPs and only has 78 of Lee’s ~10k EA SNPs.

    So with the files they distributed it does not look like it is possible to evaluate any IQ/EA data. Since the files supplied were for height it may be possible to look at earlier height GWAS results to check how their SNP results compare to the 531 SNP version used in the R code. It is necessary to have the SNP effect sizes from the GWAS though.

    Read More
    • Replies: @Factorize
    res, you see the new link that has recently been posted to an infoproc thread about genetics?
    Amazing the research is already running analysis on the Nature Genetics article, even before it is published! Something surprising from the below url, click through on the hyperlink for the Nature Genetics article. They mention that IRB only will authorize 10K of SNPs to be published. Will this apply when there are more than 10k of variants found? It seems that the rules are getting in the way of science. How exactly is one supposed to use summary statistics to actually identify anyone?
    https://www.ssc.wisc.edu/wlsresearch/documentation/GWAS/
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  41. So Piffer and Racimo seem to more or less agree than you don’t need to know all the alleles that affect a complex phenotypic trait like height or educational attainment, you just need to know some of them to see if there is positive or negative selection pressure for that trait.

    For example, Racimo sees strong selection pressure for educational achievement among East Asians, which doesn’t sound too surprising. (But it’s widespread enough in various fringe peoples, not just the Han Chinese, that it might not be, say, a product of the Mandarin exam system, but of something more ancient, like the invention of agriculture in the area.)

    On the other hand, there can be similar selection pressures with dissimilar results. Sherpas, Peruvian Indians, and highland Ethiopians have all been under selection pressures to develop adaptations to high altitude, but they apparently came up with three separate packages that work somewhat differently. This may be a literal version of the usual 3d graph about local peaks and valleys of fitness. You and your Sherpa high altitude traits can’t get to, say, Ethiopia without traveling through a lot of lowlands, probably for a few generations at least until recently, during which you high-altitude genes might prove dysfunctional.

    Read More
    • Replies: @James Thompson
    good points. Like the idea of the valleys. Rather like carrying expedition kit for all possible terrains, and dumping some of it, because it seems useless and imposes too much weight, when you stay a little too long in benign climates.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  42. Factorize says:

    Yes, this is a fascinating question: What was the driving environmental variable that selected for what we now label as educational attainment. When you go back 35,000 years the people who would later migrate to North America did not receive the genetic endowment for EA as East Asians did. It is something that happened perhaps 10-20,000 years ago (especially in East Asia). But what? By unraveling the genetic networks involved they might be able to find an answer. It might not be what we expect. In a Nordic study, it was noted that many of the students needed to travel upwards of 10 kilometers to reach their local high school especially during the 1950s. It is entirely possible that for them significant genetic variants associated with educational attainment might very well have to do with leg strength. Other variants as well might have very little connection with our expectations of a purely intellectual connection.

    Read More
    • Replies: @FKA Max

    Other variants as well might have very little connection with our expectations of a purely intellectual connection.
     
    Bingo!

    That is exactly what I was thinking, and trying to convey:

    What are Piffer et al.’s SNPs actually measuring[/indicating]?
     
    - http://www.unz.com/jthompson/piffers-equation-further-updated/#comment-2301523

    How Jews Became Smart: Anti-”Natural History of Ashkenazi Intelligence”
    R. Brian Ferguson Department of Sociology and Anthropology, Rutgers-Newark

    http://www.ncas.rutgers.edu/sites/fasn/files/How%20Jews%20Became%20Smart%20(2008).pdf

    What about those at the pinnacle, did they need high IQ’s? No doubt, it took cunning to see good opportunities. But other personality factors besides intelligence could lead to fortune.
    One could even keep this with a psychological Darwinian orientation by suggesting that risk taking, or aggressiveness-both traits often claimed to have genetic bases-led to great profit.
    Yet more then any individual qualities, the most important factors leading to greater financial
    success were possession of capital, social connections, and political patrons. And let us not forget luck–circumstances that lead to a huge payoff, or sudden ruination.
    – p. 35

    How Social Darwinism Made Modern China
    A thousand years of meritocracy shaped the Middle Kingdom.

    It seems that very harsh and “Social Darwinist” environments and cultures do not necessarily select for intelligence exclusively, or maybe not even predominately, but that they mostly select for risk taking, aggressiveness, ruthlessness, i.e., psychopathic and rather anti-social traits. - https://www.unz.com/runz/how-social-darwinism-made-modern-china-248/#comment-1866207

    Maybe what we are really looking at are "persistence/perseverance/determination" genes, not necessarily/exclusively "intelligence" genes. Confucian/Shinto, etc. genes, so to speak, that were selected for to produce conformist and ritualistic traits and personality types.

    Furthermore, the temperament trait Persistence was significantly related to some of the cognitive ability scales: Verbal Comprehension, Perceptual Reasoning, Working Memory and Full Scale IQ. This finding is in line with previous research showing that high Persistence, which is involved in the modulation of social behaviour and the control of mood and motivational drive (Gusnard et al., 2003) and effort (Cloninger et al., 1994), defines individuals who are industrious and persevering. [...] Adolescents high in Persistence are described as hardworking, and stable despite frustration and fatigue. They are also expected to increase their efforts in response to anticipated reward (Garcia, Kerekes & Archer, 2012). In other words, frustration and fatigue may be perceived as a personal challenge, they do not give up easily and are probably willing to make major sacrifices to be a success (e.g., good grades) (Garcia, Kerekes & Archer, 2012).
     
    - https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4548492/

    Explaining Asian Americans’ academic advantage over whites


    These studies show that qualities such as attentiveness, self-control, motivation, and persistencemay be as important as cognitive abilities in positively affecting academic performance.
     
    - http://www.pnas.org/content/111/23/8416

    The results show that Asian Americans rate ∼0.3 SDs lower than whites in terms of positive feelings toward themselves. They report spending ∼0.3 SDs less time with friends than their white peers. They also have more conflict with both parents than comparable white peers.

    http://www.pnas.org/content/pnas/111/23/8416/F7.medium.gif
    , @James Thompson
    Grip strength correlates with IQ
    , @Steve Sailer
    Perhaps rice cultivation selects for educational attainment?

    Or maybe millet cultivation on the Yellow River ...
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  43. Factorize says:
    @res
    I downloaded their R code to work on this. I was able to partially replicate their R results (I cut down the MCMC steps by a factor of 500, which still took me about 10 minutes to run). The results from that directory are a very small subset of what was presented in the paper.

    One thing I was not able to figure out is their code loads the parallel package, but does not appear to support multicore CPUs in any way I could see. Sorting that out would be a big help for trying a full MCMC run.

    They only included the graph format of the last plot above so I was unable to look at the other representations.

    The major problem is that their analysis requires information about the "number of ancestral and derived alleles in each population that is a leaf in our graph." Since they supplied a neutral (wrt phenotype in question) input file I had hoped to be able to use that file as a source for other SNP data. Unfortunately, it looks like they trimmed that file (for all phenotypes they looked at, or ?) so it does not have any of Piffer's 9 IQ/EA SNPs and only has 78 of Lee's ~10k EA SNPs.

    So with the files they distributed it does not look like it is possible to evaluate any IQ/EA data. Since the files supplied were for height it may be possible to look at earlier height GWAS results to check how their SNP results compare to the 531 SNP version used in the R code. It is necessary to have the SNP effect sizes from the GWAS though.

    res, you see the new link that has recently been posted to an infoproc thread about genetics?
    Amazing the research is already running analysis on the Nature Genetics article, even before it is published! Something surprising from the below url, click through on the hyperlink for the Nature Genetics article. They mention that IRB only will authorize 10K of SNPs to be published. Will this apply when there are more than 10k of variants found? It seems that the rules are getting in the way of science. How exactly is one supposed to use summary statistics to actually identify anyone?

    https://www.ssc.wisc.edu/wlsresearch/documentation/GWAS/

    Read More
    • Replies: @res
    There is definitely some excitement over the upcoming Lee paper!

    The WLS was new to me. Thanks.

    How exactly is one supposed to use summary statistics to actually identify anyone?
     
    Using a subset of SNPs to identify someone seems quite possible. Especially in conjunction with even basic data like age.

    Some discussion: https://www.biostars.org/p/167494/
    This is about relatedness, but the links might be of use: https://www.researchgate.net/post/How_many_SNP_markers_are_generally_required_to_identify_kinship_in_the_first_second_third_degree

    The forensics literature is probably the place to go for this information. For example: https://academic.oup.com/bioscience/article/58/6/484/235866

    Not SNPs, but useful background:

    The standard DNA profile collected in the United States and entered into CODIS consists of 13 STR loci plus the amelogenin gene, which is found on the X and Y chromosomes and can establish the sex of unknown sample sources. The probability that two unrelated individuals share the total profile is less than one in one trillion. CODIS currently contains nearly 6 million STR profiles

     

    This makes one think:


    The primary disadvantage of SNPs is that they do not exist in as many different varieties as do STRs, and therefore their power for making unique identifications is considerably less. Approximately 50 SNPs are required to identify an individual with the same certainty as with the standard 13 STRs.
     
    I find that 50 number hard to believe unless they are specifically selecting high variability SNPs.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  44. res says:
    @Factorize
    res, you see the new link that has recently been posted to an infoproc thread about genetics?
    Amazing the research is already running analysis on the Nature Genetics article, even before it is published! Something surprising from the below url, click through on the hyperlink for the Nature Genetics article. They mention that IRB only will authorize 10K of SNPs to be published. Will this apply when there are more than 10k of variants found? It seems that the rules are getting in the way of science. How exactly is one supposed to use summary statistics to actually identify anyone?
    https://www.ssc.wisc.edu/wlsresearch/documentation/GWAS/

    There is definitely some excitement over the upcoming Lee paper!

    The WLS was new to me. Thanks.

    How exactly is one supposed to use summary statistics to actually identify anyone?

    Using a subset of SNPs to identify someone seems quite possible. Especially in conjunction with even basic data like age.

    Some discussion: https://www.biostars.org/p/167494/
    This is about relatedness, but the links might be of use: https://www.researchgate.net/post/How_many_SNP_markers_are_generally_required_to_identify_kinship_in_the_first_second_third_degree

    The forensics literature is probably the place to go for this information. For example: https://academic.oup.com/bioscience/article/58/6/484/235866

    Not SNPs, but useful background:

    The standard DNA profile collected in the United States and entered into CODIS consists of 13 STR loci plus the amelogenin gene, which is found on the X and Y chromosomes and can establish the sex of unknown sample sources. The probability that two unrelated individuals share the total profile is less than one in one trillion. CODIS currently contains nearly 6 million STR profiles

    This makes one think:

    The primary disadvantage of SNPs is that they do not exist in as many different varieties as do STRs, and therefore their power for making unique identifications is considerably less. Approximately 50 SNPs are required to identify an individual with the same certainty as with the standard 13 STRs.

    I find that 50 number hard to believe unless they are specifically selecting high variability SNPs.

    Read More
    • Replies: @Factorize
    It just occurred to me right now: Compressed sensing or other method that considers all of the EA SNPs in a global optimization method. Yeah!! That would be awesome. I suppose there is some very big computer crunching through those SNPs right about now. How much more of the picture should that give us? Not sure why the authors didn't try that themselves. They are leaving quite a bit on the table for others to claim. We could have a whole round of first and second generation replies to the Nature Genetics article before it even makes it online. It was a great idea to leak the details because now everyone is biting their fingernails.
    , @Factorize
    res, 2^50 should do it ~10^15.

    I have been thinking about how organizing genomes will prevent regression to the mean from occurring. As it is now, some fortunate individuals can hit the genetic jackpot and receive especially
    combinations of chromosomal strands and chromosomal recombinants. However, as this is a random process, there is a very high likelihood that such good fortune will not be maintained in future generations: regression to the mean.

    What happens, though, with embryo selection when genomes could be organized in such a way to prevent this from ever occurring again? As an example, suppose a person has a high polygenic score on at least one strand of chromosomes 1 and 2. A partner could be found with this same pattern and embryo selection could be used such that the embryo would have high PGS on both strands. When this child became an adult, only high PGS gametes on chromosomes 1 and 2 would be produced. This process could be repeated and each generation would fix at least 2 more chromosomes. A very large, near maximal PGS would emerge after 5 generations.

    The process is still random, though the dice are now fixed. You win no matter what happens. Regression to the mean has been avoided. In each generation you could fix a minimum of 2 chromosomes using 16 fold embryo selection. However, of the 23 chromosomes, several might have at least 1 strong strand, so there might be greater than 2 fixations that occur per generation.

    The veil of ignorance would be lifted. This line of thinking has substantial social consequences.
    Until now, we have all been subject to a random genetic process: The long term outcome of which is to return inevitably to average. With regression to the mean avoided, rational self-interest suggests that people would focus to a greater extent on their personal (instead of collective) utility.

    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  45. FKA Max says: • Website
    @Factorize
    Yes, this is a fascinating question: What was the driving environmental variable that selected for what we now label as educational attainment. When you go back 35,000 years the people who would later migrate to North America did not receive the genetic endowment for EA as East Asians did. It is something that happened perhaps 10-20,000 years ago (especially in East Asia). But what? By unraveling the genetic networks involved they might be able to find an answer. It might not be what we expect. In a Nordic study, it was noted that many of the students needed to travel upwards of 10 kilometers to reach their local high school especially during the 1950s. It is entirely possible that for them significant genetic variants associated with educational attainment might very well have to do with leg strength. Other variants as well might have very little connection with our expectations of a purely intellectual connection.

    Other variants as well might have very little connection with our expectations of a purely intellectual connection.

    Bingo!

    That is exactly what I was thinking, and trying to convey:

    What are Piffer et al.’s SNPs actually measuring[/indicating]?

    http://www.unz.com/jthompson/piffers-equation-further-updated/#comment-2301523

    How Jews Became Smart: Anti-”Natural History of Ashkenazi Intelligence”
    R. Brian Ferguson Department of Sociology and Anthropology, Rutgers-Newark

    http://www.ncas.rutgers.edu/sites/fasn/files/How%20Jews%20Became%20Smart%20(2008).pdf

    What about those at the pinnacle, did they need high IQ’s? No doubt, it took cunning to see good opportunities. But other personality factors besides intelligence could lead to fortune.
    One could even keep this with a psychological Darwinian orientation by suggesting that risk taking, or aggressiveness-both traits often claimed to have genetic bases-led to great profit.
    Yet more then any individual qualities, the most important factors leading to greater financial
    success were possession of capital, social connections, and political patrons. And let us not forget luck–circumstances that lead to a huge payoff, or sudden ruination.
    – p. 35

    How Social Darwinism Made Modern China
    A thousand years of meritocracy shaped the Middle Kingdom.

    It seems that very harsh and “Social Darwinist” environments and cultures do not necessarily select for intelligence exclusively, or maybe not even predominately, but that they mostly select for risk taking, aggressiveness, ruthlessness, i.e., psychopathic and rather anti-social traits.https://www.unz.com/runz/how-social-darwinism-made-modern-china-248/#comment-1866207

    Maybe what we are really looking at are “persistence/perseverance/determination” genes, not necessarily/exclusively “intelligence” genes. Confucian/Shinto, etc. genes, so to speak, that were selected for to produce conformist and ritualistic traits and personality types.

    Furthermore, the temperament trait Persistence was significantly related to some of the cognitive ability scales: Verbal Comprehension, Perceptual Reasoning, Working Memory and Full Scale IQ. This finding is in line with previous research showing that high Persistence, which is involved in the modulation of social behaviour and the control of mood and motivational drive (Gusnard et al., 2003) and effort (Cloninger et al., 1994), defines individuals who are industrious and persevering. [...] Adolescents high in Persistence are described as hardworking, and stable despite frustration and fatigue. They are also expected to increase their efforts in response to anticipated reward (Garcia, Kerekes & Archer, 2012). In other words, frustration and fatigue may be perceived as a personal challenge, they do not give up easily and are probably willing to make major sacrifices to be a success (e.g., good grades) (Garcia, Kerekes & Archer, 2012).

    https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4548492/

    Explaining Asian Americans’ academic advantage over whites

    These studies show that qualities such as attentiveness, self-control, motivation, and persistencemay be as important as cognitive abilities in positively affecting academic performance.

    http://www.pnas.org/content/111/23/8416

    The results show that Asian Americans rate ∼0.3 SDs lower than whites in terms of positive feelings toward themselves. They report spending ∼0.3 SDs less time with friends than their white peers. They also have more conflict with both parents than comparable white peers.

    Read More
    • Replies: @James Thompson
    WD Furneaux . Speed, accuracy, and Persistence. The last, a neglected aspect of intellect.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  46. @Steve Sailer
    So Piffer and Racimo seem to more or less agree than you don't need to know all the alleles that affect a complex phenotypic trait like height or educational attainment, you just need to know some of them to see if there is positive or negative selection pressure for that trait.

    For example, Racimo sees strong selection pressure for educational achievement among East Asians, which doesn't sound too surprising. (But it's widespread enough in various fringe peoples, not just the Han Chinese, that it might not be, say, a product of the Mandarin exam system, but of something more ancient, like the invention of agriculture in the area.)

    On the other hand, there can be similar selection pressures with dissimilar results. Sherpas, Peruvian Indians, and highland Ethiopians have all been under selection pressures to develop adaptations to high altitude, but they apparently came up with three separate packages that work somewhat differently. This may be a literal version of the usual 3d graph about local peaks and valleys of fitness. You and your Sherpa high altitude traits can't get to, say, Ethiopia without traveling through a lot of lowlands, probably for a few generations at least until recently, during which you high-altitude genes might prove dysfunctional.

    good points. Like the idea of the valleys. Rather like carrying expedition kit for all possible terrains, and dumping some of it, because it seems useless and imposes too much weight, when you stay a little too long in benign climates.

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  47. @Factorize
    Yes, this is a fascinating question: What was the driving environmental variable that selected for what we now label as educational attainment. When you go back 35,000 years the people who would later migrate to North America did not receive the genetic endowment for EA as East Asians did. It is something that happened perhaps 10-20,000 years ago (especially in East Asia). But what? By unraveling the genetic networks involved they might be able to find an answer. It might not be what we expect. In a Nordic study, it was noted that many of the students needed to travel upwards of 10 kilometers to reach their local high school especially during the 1950s. It is entirely possible that for them significant genetic variants associated with educational attainment might very well have to do with leg strength. Other variants as well might have very little connection with our expectations of a purely intellectual connection.

    Grip strength correlates with IQ

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  48. @Factorize
    Yes, this is a fascinating question: What was the driving environmental variable that selected for what we now label as educational attainment. When you go back 35,000 years the people who would later migrate to North America did not receive the genetic endowment for EA as East Asians did. It is something that happened perhaps 10-20,000 years ago (especially in East Asia). But what? By unraveling the genetic networks involved they might be able to find an answer. It might not be what we expect. In a Nordic study, it was noted that many of the students needed to travel upwards of 10 kilometers to reach their local high school especially during the 1950s. It is entirely possible that for them significant genetic variants associated with educational attainment might very well have to do with leg strength. Other variants as well might have very little connection with our expectations of a purely intellectual connection.

    Perhaps rice cultivation selects for educational attainment?

    Or maybe millet cultivation on the Yellow River …

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

    It might just be the idea more broadly of resource constraints that existed in East Asia. Imagine the resources that would have been available to the first settlers of the Americas! It must have felt like an almost limitless frontier.

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter Display All Comments
  50. Factorize says:
    @res
    There is definitely some excitement over the upcoming Lee paper!

    The WLS was new to me. Thanks.

    How exactly is one supposed to use summary statistics to actually identify anyone?
     
    Using a subset of SNPs to identify someone seems quite possible. Especially in conjunction with even basic data like age.

    Some discussion: https://www.biostars.org/p/167494/
    This is about relatedness, but the links might be of use: https://www.researchgate.net/post/How_many_SNP_markers_are_generally_required_to_identify_kinship_in_the_first_second_third_degree

    The forensics literature is probably the place to go for this information. For example: https://academic.oup.com/bioscience/article/58/6/484/235866

    Not SNPs, but useful background:

    The standard DNA profile collected in the United States and entered into CODIS consists of 13 STR loci plus the amelogenin gene, which is found on the X and Y chromosomes and can establish the sex of unknown sample sources. The probability that two unrelated individuals share the total profile is less than one in one trillion. CODIS currently contains nearly 6 million STR profiles

     

    This makes one think:


    The primary disadvantage of SNPs is that they do not exist in as many different varieties as do STRs, and therefore their power for making unique identifications is considerably less. Approximately 50 SNPs are required to identify an individual with the same certainty as with the standard 13 STRs.
     
    I find that 50 number hard to believe unless they are specifically selecting high variability SNPs.

    It just occurred to me right now: Compressed sensing or other method that considers all of the EA SNPs in a global optimization method. Yeah!! That would be awesome. I suppose there is some very big computer crunching through those SNPs right about now. How much more of the picture should that give us? Not sure why the authors didn’t try that themselves. They are leaving quite a bit on the table for others to claim. We could have a whole round of first and second generation replies to the Nature Genetics article before it even makes it online. It was a great idea to leak the details because now everyone is biting their fingernails.

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  51. @FKA Max

    Other variants as well might have very little connection with our expectations of a purely intellectual connection.
     
    Bingo!

    That is exactly what I was thinking, and trying to convey:

    What are Piffer et al.’s SNPs actually measuring[/indicating]?
     
    - http://www.unz.com/jthompson/piffers-equation-further-updated/#comment-2301523

    How Jews Became Smart: Anti-”Natural History of Ashkenazi Intelligence”
    R. Brian Ferguson Department of Sociology and Anthropology, Rutgers-Newark

    http://www.ncas.rutgers.edu/sites/fasn/files/How%20Jews%20Became%20Smart%20(2008).pdf

    What about those at the pinnacle, did they need high IQ’s? No doubt, it took cunning to see good opportunities. But other personality factors besides intelligence could lead to fortune.
    One could even keep this with a psychological Darwinian orientation by suggesting that risk taking, or aggressiveness-both traits often claimed to have genetic bases-led to great profit.
    Yet more then any individual qualities, the most important factors leading to greater financial
    success were possession of capital, social connections, and political patrons. And let us not forget luck–circumstances that lead to a huge payoff, or sudden ruination.
    – p. 35

    How Social Darwinism Made Modern China
    A thousand years of meritocracy shaped the Middle Kingdom.

    It seems that very harsh and “Social Darwinist” environments and cultures do not necessarily select for intelligence exclusively, or maybe not even predominately, but that they mostly select for risk taking, aggressiveness, ruthlessness, i.e., psychopathic and rather anti-social traits. - https://www.unz.com/runz/how-social-darwinism-made-modern-china-248/#comment-1866207

    Maybe what we are really looking at are "persistence/perseverance/determination" genes, not necessarily/exclusively "intelligence" genes. Confucian/Shinto, etc. genes, so to speak, that were selected for to produce conformist and ritualistic traits and personality types.

    Furthermore, the temperament trait Persistence was significantly related to some of the cognitive ability scales: Verbal Comprehension, Perceptual Reasoning, Working Memory and Full Scale IQ. This finding is in line with previous research showing that high Persistence, which is involved in the modulation of social behaviour and the control of mood and motivational drive (Gusnard et al., 2003) and effort (Cloninger et al., 1994), defines individuals who are industrious and persevering. [...] Adolescents high in Persistence are described as hardworking, and stable despite frustration and fatigue. They are also expected to increase their efforts in response to anticipated reward (Garcia, Kerekes & Archer, 2012). In other words, frustration and fatigue may be perceived as a personal challenge, they do not give up easily and are probably willing to make major sacrifices to be a success (e.g., good grades) (Garcia, Kerekes & Archer, 2012).
     
    - https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4548492/

    Explaining Asian Americans’ academic advantage over whites


    These studies show that qualities such as attentiveness, self-control, motivation, and persistencemay be as important as cognitive abilities in positively affecting academic performance.
     
    - http://www.pnas.org/content/111/23/8416

    The results show that Asian Americans rate ∼0.3 SDs lower than whites in terms of positive feelings toward themselves. They report spending ∼0.3 SDs less time with friends than their white peers. They also have more conflict with both parents than comparable white peers.

    http://www.pnas.org/content/pnas/111/23/8416/F7.medium.gif

    WD Furneaux . Speed, accuracy, and Persistence. The last, a neglected aspect of intellect.

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  52. Factorize says:
    @res
    There is definitely some excitement over the upcoming Lee paper!

    The WLS was new to me. Thanks.

    How exactly is one supposed to use summary statistics to actually identify anyone?
     
    Using a subset of SNPs to identify someone seems quite possible. Especially in conjunction with even basic data like age.

    Some discussion: https://www.biostars.org/p/167494/
    This is about relatedness, but the links might be of use: https://www.researchgate.net/post/How_many_SNP_markers_are_generally_required_to_identify_kinship_in_the_first_second_third_degree

    The forensics literature is probably the place to go for this information. For example: https://academic.oup.com/bioscience/article/58/6/484/235866

    Not SNPs, but useful background:

    The standard DNA profile collected in the United States and entered into CODIS consists of 13 STR loci plus the amelogenin gene, which is found on the X and Y chromosomes and can establish the sex of unknown sample sources. The probability that two unrelated individuals share the total profile is less than one in one trillion. CODIS currently contains nearly 6 million STR profiles

     

    This makes one think:


    The primary disadvantage of SNPs is that they do not exist in as many different varieties as do STRs, and therefore their power for making unique identifications is considerably less. Approximately 50 SNPs are required to identify an individual with the same certainty as with the standard 13 STRs.
     
    I find that 50 number hard to believe unless they are specifically selecting high variability SNPs.

    res, 2^50 should do it ~10^15.

    I have been thinking about how organizing genomes will prevent regression to the mean from occurring. As it is now, some fortunate individuals can hit the genetic jackpot and receive especially
    combinations of chromosomal strands and chromosomal recombinants. However, as this is a random process, there is a very high likelihood that such good fortune will not be maintained in future generations: regression to the mean.

    What happens, though, with embryo selection when genomes could be organized in such a way to prevent this from ever occurring again? As an example, suppose a person has a high polygenic score on at least one strand of chromosomes 1 and 2. A partner could be found with this same pattern and embryo selection could be used such that the embryo would have high PGS on both strands. When this child became an adult, only high PGS gametes on chromosomes 1 and 2 would be produced. This process could be repeated and each generation would fix at least 2 more chromosomes. A very large, near maximal PGS would emerge after 5 generations.

    The process is still random, though the dice are now fixed. You win no matter what happens. Regression to the mean has been avoided. In each generation you could fix a minimum of 2 chromosomes using 16 fold embryo selection. However, of the 23 chromosomes, several might have at least 1 strong strand, so there might be greater than 2 fixations that occur per generation.

    The veil of ignorance would be lifted. This line of thinking has substantial social consequences.
    Until now, we have all been subject to a random genetic process: The long term outcome of which is to return inevitably to average. With regression to the mean avoided, rational self-interest suggests that people would focus to a greater extent on their personal (instead of collective) utility.

    Read More
    • Replies: @res

    res, 2^50 should do it ~10^15.
     
    I guess so. As long as reasonably variable SNPs are used. Especially since a single SNP (unphased) has three states: aa Aa AA.

    The aspects I'm not sure how to estimate are how MAF affects the resolving power of the SNPs and how systematic variations (family, racial, etc.) affect the match probabilities. If we think of the SNPs as binary we can use information theory entropy to estimate the number of bits of information supplied by a particular SNP: https://en.wikipedia.org/wiki/Entropy_(information_theory)#Definition

    For two equally probable states we have : -(1/2 * log2(1/2))*2 = 1 bit
    For 25/75 allele frequencies we have - (1/4*log2(1/4) + 3/4*log2(3/4)) = - (-0.5 + -0.31) = 0.81
    For 1/99 AF we have -(1/100*log2(1/100) + 99/100*log2(99/100)) = -(-0.006 + -0.014) = 0.02

    So it is clear that the variability of the 50 SNPs in the population is critical to their resolving power. Here is a plot of the binary case from that wiki link:

    https://upload.wikimedia.org/wikipedia/commons/thumb/2/22/Binary_entropy_plot.svg/300px-Binary_entropy_plot.svg.png

    From that we see that as long as the MAF is > ~0.10 we have decent resolving power (over a half bit), but it drops sharply below that MAF.

    A similar analysis can be done for the 3 state frequencies.

    embryo selection could be used such that the embryo would have high PGS on both strands.
     
    What about recombination? I think your approach requires a large population in the second generation OR multiple first generation groups fixing different chromosomes. It would be interesting to see how a simulation of your approach compares to brute force approaches with different numbers of generations.

    With regression to the mean avoided, rational self-interest suggests that people would focus to a greater extent on their personal (instead of collective) utility.
     
    That is a very interesting (and rather frightening) thought. How does that differ from the existing case of different groups regressing to rather different means?
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  53. res says:
    @Factorize
    res, 2^50 should do it ~10^15.

    I have been thinking about how organizing genomes will prevent regression to the mean from occurring. As it is now, some fortunate individuals can hit the genetic jackpot and receive especially
    combinations of chromosomal strands and chromosomal recombinants. However, as this is a random process, there is a very high likelihood that such good fortune will not be maintained in future generations: regression to the mean.

    What happens, though, with embryo selection when genomes could be organized in such a way to prevent this from ever occurring again? As an example, suppose a person has a high polygenic score on at least one strand of chromosomes 1 and 2. A partner could be found with this same pattern and embryo selection could be used such that the embryo would have high PGS on both strands. When this child became an adult, only high PGS gametes on chromosomes 1 and 2 would be produced. This process could be repeated and each generation would fix at least 2 more chromosomes. A very large, near maximal PGS would emerge after 5 generations.

    The process is still random, though the dice are now fixed. You win no matter what happens. Regression to the mean has been avoided. In each generation you could fix a minimum of 2 chromosomes using 16 fold embryo selection. However, of the 23 chromosomes, several might have at least 1 strong strand, so there might be greater than 2 fixations that occur per generation.

    The veil of ignorance would be lifted. This line of thinking has substantial social consequences.
    Until now, we have all been subject to a random genetic process: The long term outcome of which is to return inevitably to average. With regression to the mean avoided, rational self-interest suggests that people would focus to a greater extent on their personal (instead of collective) utility.

    res, 2^50 should do it ~10^15.

    I guess so. As long as reasonably variable SNPs are used. Especially since a single SNP (unphased) has three states: aa Aa AA.

    The aspects I’m not sure how to estimate are how MAF affects the resolving power of the SNPs and how systematic variations (family, racial, etc.) affect the match probabilities. If we think of the SNPs as binary we can use information theory entropy to estimate the number of bits of information supplied by a particular SNP: https://en.wikipedia.org/wiki/Entropy_(information_theory)#Definition

    For two equally probable states we have : -(1/2 * log2(1/2))*2 = 1 bit
    For 25/75 allele frequencies we have – (1/4*log2(1/4) + 3/4*log2(3/4)) = – (-0.5 + -0.31) = 0.81
    For 1/99 AF we have -(1/100*log2(1/100) + 99/100*log2(99/100)) = -(-0.006 + -0.014) = 0.02

    So it is clear that the variability of the 50 SNPs in the population is critical to their resolving power. Here is a plot of the binary case from that wiki link:

    From that we see that as long as the MAF is > ~0.10 we have decent resolving power (over a half bit), but it drops sharply below that MAF.

    A similar analysis can be done for the 3 state frequencies.

    embryo selection could be used such that the embryo would have high PGS on both strands.

    What about recombination? I think your approach requires a large population in the second generation OR multiple first generation groups fixing different chromosomes. It would be interesting to see how a simulation of your approach compares to brute force approaches with different numbers of generations.

    With regression to the mean avoided, rational self-interest suggests that people would focus to a greater extent on their personal (instead of collective) utility.

    That is a very interesting (and rather frightening) thought. How does that differ from the existing case of different groups regressing to rather different means?

    Read More
    • Replies: @Factorize
    Good point about the 3 genotype states. 50 SNPs seems excessive, though as you noted there likely could be genetic patterns that could hinder the analysis. Thank you for including the discussion about the information theory discussion! I had not encountered this line of thought before.

    For organizing of the genome through generations, I would expect that job 1 would be do a few rounds of sorting out the high and low PGS chromosomes. This might take 5 or more generations to accomplish. After that point, further maximizing of the PGS could look at the recombinants. It is amazing how much of the genetic uplift in human intelligence will result from very basic reproductive technology. I suppose there are a number of people who are not entirely amused by this. It will be only after many rounds of embryo selection will we finally reach the point where gene editing would be competitive. One interesting and somewhat neglected aspect of this discussion that has been lost is that with an extra X chromosome (which is a fairly substantial chromosome), women would be expected to gain a psychometric edge over men as the Y Chromosome has greatly atrophied through time and probably does not contain the same amount of psychometric potential as the X.

    Large mating pools would be a helpful, though not an overly necessary feature of the mating scheme that I proposed. In this respect China will have an enormous advantage over other nations.
    However, it will require that the population could move towards a more homegenous and single mating population. Being able to find a partner with the most complementary DNA would create enormous advantages. In the scheme that I propose it is no longer strictly necessary to think in terms of looking for a partner with the highest possible PGS score. What one is really looking for is the best complementary DNA. It probably could make a great deal of sense if one had an especially high PGS on one chromosome to continue to select for this chromosome through time. This would take one shot on goal. For the numbers that I saw, there might really only be 7-8 chromosomes that drive the PGS score. To maintain a high score, one would want to become part of a mating pool that had very high values for these chromosomes.

    Yes, you are right, at least initially different mating pools who were trying to maximize different sets of chromosomes would probably form. Some might be interested in fixing Chromosomes 1 and 2, others 3 and 4. However, after 3 or 4 generations a community scale mating pool would then be re-established. Some might even accept hitting a jackpot with only a very high PGS without there being any increased genomic order. It will be fascinating to see what choices people make regarding short term or long term benefits. How many IQ points would someone forego now in order to be certain of fixing 2 chromosomes now?

    The idea of personal versus collective utility is scary; I tried to word it carefully. To be less obscure, what role will the good of the collective mean for those who can be assured that their family will never return to this collective? Until the recent research, I would fully endorse this broad conception of a shared commons because random assortment has essentially ensured that after a few generations will all return to such an average (though as you noted, this average is different for different groups). Knowing that this is no longer a certainty clearly could mean that the sense that we are all in this together could fade. Genetics technology will mean that regression to the mean will not occur again. We are not all in this together.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  54. Factorize says:
    @res

    res, 2^50 should do it ~10^15.
     
    I guess so. As long as reasonably variable SNPs are used. Especially since a single SNP (unphased) has three states: aa Aa AA.

    The aspects I'm not sure how to estimate are how MAF affects the resolving power of the SNPs and how systematic variations (family, racial, etc.) affect the match probabilities. If we think of the SNPs as binary we can use information theory entropy to estimate the number of bits of information supplied by a particular SNP: https://en.wikipedia.org/wiki/Entropy_(information_theory)#Definition

    For two equally probable states we have : -(1/2 * log2(1/2))*2 = 1 bit
    For 25/75 allele frequencies we have - (1/4*log2(1/4) + 3/4*log2(3/4)) = - (-0.5 + -0.31) = 0.81
    For 1/99 AF we have -(1/100*log2(1/100) + 99/100*log2(99/100)) = -(-0.006 + -0.014) = 0.02

    So it is clear that the variability of the 50 SNPs in the population is critical to their resolving power. Here is a plot of the binary case from that wiki link:

    https://upload.wikimedia.org/wikipedia/commons/thumb/2/22/Binary_entropy_plot.svg/300px-Binary_entropy_plot.svg.png

    From that we see that as long as the MAF is > ~0.10 we have decent resolving power (over a half bit), but it drops sharply below that MAF.

    A similar analysis can be done for the 3 state frequencies.

    embryo selection could be used such that the embryo would have high PGS on both strands.
     
    What about recombination? I think your approach requires a large population in the second generation OR multiple first generation groups fixing different chromosomes. It would be interesting to see how a simulation of your approach compares to brute force approaches with different numbers of generations.

    With regression to the mean avoided, rational self-interest suggests that people would focus to a greater extent on their personal (instead of collective) utility.
     
    That is a very interesting (and rather frightening) thought. How does that differ from the existing case of different groups regressing to rather different means?

    Good point about the 3 genotype states. 50 SNPs seems excessive, though as you noted there likely could be genetic patterns that could hinder the analysis. Thank you for including the discussion about the information theory discussion! I had not encountered this line of thought before.

    For organizing of the genome through generations, I would expect that job 1 would be do a few rounds of sorting out the high and low PGS chromosomes. This might take 5 or more generations to accomplish. After that point, further maximizing of the PGS could look at the recombinants. It is amazing how much of the genetic uplift in human intelligence will result from very basic reproductive technology. I suppose there are a number of people who are not entirely amused by this. It will be only after many rounds of embryo selection will we finally reach the point where gene editing would be competitive. One interesting and somewhat neglected aspect of this discussion that has been lost is that with an extra X chromosome (which is a fairly substantial chromosome), women would be expected to gain a psychometric edge over men as the Y Chromosome has greatly atrophied through time and probably does not contain the same amount of psychometric potential as the X.

    Large mating pools would be a helpful, though not an overly necessary feature of the mating scheme that I proposed. In this respect China will have an enormous advantage over other nations.
    However, it will require that the population could move towards a more homegenous and single mating population. Being able to find a partner with the most complementary DNA would create enormous advantages. In the scheme that I propose it is no longer strictly necessary to think in terms of looking for a partner with the highest possible PGS score. What one is really looking for is the best complementary DNA. It probably could make a great deal of sense if one had an especially high PGS on one chromosome to continue to select for this chromosome through time. This would take one shot on goal. For the numbers that I saw, there might really only be 7-8 chromosomes that drive the PGS score. To maintain a high score, one would want to become part of a mating pool that had very high values for these chromosomes.

    Yes, you are right, at least initially different mating pools who were trying to maximize different sets of chromosomes would probably form. Some might be interested in fixing Chromosomes 1 and 2, others 3 and 4. However, after 3 or 4 generations a community scale mating pool would then be re-established. Some might even accept hitting a jackpot with only a very high PGS without there being any increased genomic order. It will be fascinating to see what choices people make regarding short term or long term benefits. How many IQ points would someone forego now in order to be certain of fixing 2 chromosomes now?

    The idea of personal versus collective utility is scary; I tried to word it carefully. To be less obscure, what role will the good of the collective mean for those who can be assured that their family will never return to this collective? Until the recent research, I would fully endorse this broad conception of a shared commons because random assortment has essentially ensured that after a few generations will all return to such an average (though as you noted, this average is different for different groups). Knowing that this is no longer a certainty clearly could mean that the sense that we are all in this together could fade. Genetics technology will mean that regression to the mean will not occur again. We are not all in this together.

    Read More
    • Replies: @res

    One interesting and somewhat neglected aspect of this discussion that has been lost is that with an extra X chromosome (which is a fairly substantial chromosome), women would be expected to gain a psychometric edge over men as the Y Chromosome has greatly atrophied through time and probably does not contain the same amount of psychometric potential as the X.
     
    I don't think the X chromosome works that way. See https://en.wikipedia.org/wiki/X-inactivation

    Where it probably does matter is increasing the variance for males vs. females.

    Interesting last paragraph.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  55. res says:
    @Factorize
    Good point about the 3 genotype states. 50 SNPs seems excessive, though as you noted there likely could be genetic patterns that could hinder the analysis. Thank you for including the discussion about the information theory discussion! I had not encountered this line of thought before.

    For organizing of the genome through generations, I would expect that job 1 would be do a few rounds of sorting out the high and low PGS chromosomes. This might take 5 or more generations to accomplish. After that point, further maximizing of the PGS could look at the recombinants. It is amazing how much of the genetic uplift in human intelligence will result from very basic reproductive technology. I suppose there are a number of people who are not entirely amused by this. It will be only after many rounds of embryo selection will we finally reach the point where gene editing would be competitive. One interesting and somewhat neglected aspect of this discussion that has been lost is that with an extra X chromosome (which is a fairly substantial chromosome), women would be expected to gain a psychometric edge over men as the Y Chromosome has greatly atrophied through time and probably does not contain the same amount of psychometric potential as the X.

    Large mating pools would be a helpful, though not an overly necessary feature of the mating scheme that I proposed. In this respect China will have an enormous advantage over other nations.
    However, it will require that the population could move towards a more homegenous and single mating population. Being able to find a partner with the most complementary DNA would create enormous advantages. In the scheme that I propose it is no longer strictly necessary to think in terms of looking for a partner with the highest possible PGS score. What one is really looking for is the best complementary DNA. It probably could make a great deal of sense if one had an especially high PGS on one chromosome to continue to select for this chromosome through time. This would take one shot on goal. For the numbers that I saw, there might really only be 7-8 chromosomes that drive the PGS score. To maintain a high score, one would want to become part of a mating pool that had very high values for these chromosomes.

    Yes, you are right, at least initially different mating pools who were trying to maximize different sets of chromosomes would probably form. Some might be interested in fixing Chromosomes 1 and 2, others 3 and 4. However, after 3 or 4 generations a community scale mating pool would then be re-established. Some might even accept hitting a jackpot with only a very high PGS without there being any increased genomic order. It will be fascinating to see what choices people make regarding short term or long term benefits. How many IQ points would someone forego now in order to be certain of fixing 2 chromosomes now?

    The idea of personal versus collective utility is scary; I tried to word it carefully. To be less obscure, what role will the good of the collective mean for those who can be assured that their family will never return to this collective? Until the recent research, I would fully endorse this broad conception of a shared commons because random assortment has essentially ensured that after a few generations will all return to such an average (though as you noted, this average is different for different groups). Knowing that this is no longer a certainty clearly could mean that the sense that we are all in this together could fade. Genetics technology will mean that regression to the mean will not occur again. We are not all in this together.

    One interesting and somewhat neglected aspect of this discussion that has been lost is that with an extra X chromosome (which is a fairly substantial chromosome), women would be expected to gain a psychometric edge over men as the Y Chromosome has greatly atrophied through time and probably does not contain the same amount of psychometric potential as the X.

    I don’t think the X chromosome works that way. See https://en.wikipedia.org/wiki/X-inactivation

    Where it probably does matter is increasing the variance for males vs. females.

    Interesting last paragraph.

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

    res, good one res!
    X-Inactivation, of course!

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter Display All Comments
  57. Jm8 says:

    (interestingly) Even without the anomalously high scores of the Pygmies, Africans (in the above scores) generally score higher than Amerindians overall. E.g. Kenyan Bantus, South African Bantus, Yorubas, and Mandenkas score higher than Maya and Brazilian Surui Indians (and the Bantu populations and Mandenka score higher than Brazilian Surui and Brazilian Karitiana Indians, and slightly higher than-roughly the same as Pima Indians).

    Read More
    • Replies: @res
    That is definitely interesting. I suspect it is for the same reason that height PGS don't work properly for Pygmies (Pygmies are just the most extreme case we see). IIRC that is because some variants that are important for Pygmy height don't show up in European GWAS. Amerindians are closer to Europeans than Africans genetically so I could see that effect existing between Amerindians and Africans there as well.

    I think this effect would result in Piffer's PGS predicting higher IQs for Africa than the reality. If so, it is ironic that some possible issues Piffer's critics highlight might result in the true numbers being different in a direction they don't like.

    To tie this into the DUF1220 discussion elsewhere, this paper implicates DUF1220 in brain size: https://www.sciencedirect.com/science/article/pii/S0002929712003734
    They actually look at ethnic differences. They find no significant association, but given the sample sizes and p value I am suspicious.

    The ethnic distribution of individuals with known deletions or duplications was categorized as 26 white, 12 Hispanic, and 3 African American or other. Differences of copy ratio between ethnic groups were tested with a one-way ANOVA. We assessed population stratification in the group with a known deletion by including ethnicity as a covariate in a linear regression model of FOC Z score and CON1. There was no evidence of (1) a significant difference of copy ratio between ethnic groups (p > 0.10), (2) confounding by ethnicity, or (3) an association between ethnicity and FOC Z scores.
     
    Here is an old blog post looking at this: http://dienekes.blogspot.com/2012/08/our-big-human-brains-may-depend-on.html
    The cross species correlation is dramatic! And the human/Neandertal difference is very interesting.

    http://2.bp.blogspot.com/-tTV5L6PVMg4/UC0-6nXIKvI/AAAAAAAAFa8/dzu_4yBjwR8/s320/brain.jpg

    This paper does not cover DUF1220 (AFAICT), but gives an idea of the kind of CNV analysis which was possible in 2010: http://science.sciencemag.org/content/330/6004/641.full

    https://d2ufo47lrtsv5s.cloudfront.net/content/sci/330/6004/641/F3.large.jpg

    The author's later paper looks at DUF1220 in the great ape lineage, but not intra-human differences: https://genome.cshlp.org/content/23/9/1373.full

    This 2015 paper by the same author looks at intra-human structural variations, but no mention of DUF1220 (interesting when dogs don't bark): https://www.nature.com/articles/nature15394

    Figure S2B here: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5556342/bin/12864_2017_3976_MOESM2_ESM.pdf
    From this paper: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5556342/
    Shows boxplots by 1000 Genomes population for different areas of DUF1220 variation. The groups have somewhat differing means, but overlap heavily. Worth noting that the ASW population had high means. That figure looks like evidence to me that there are NOT systematic significant differences in DUF1220 copies by race.

    Worth repeating from the paper discussed in another comment: https://link.springer.com/article/10.1007%2Fs00439-014-1489-2

    Given these fndings, we examined associations between DUF1220 subtypes CON1 and CON2 and cognitive aptitude. We identifed a linear association between CON2 copy number and cognitive function in two independent populations of European descent.

     

    So CON2 is probably the region to focus on. Unfortunately there are relatively few copies there so Figure S2B is not really readable in that case.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  58. res says:
    @Jm8
    (interestingly) Even without the anomalously high scores of the Pygmies, Africans (in the above scores) generally score higher than Amerindians overall. E.g. Kenyan Bantus, South African Bantus, Yorubas, and Mandenkas score higher than Maya and Brazilian Surui Indians (and the Bantu populations and Mandenka score higher than Brazilian Surui and Brazilian Karitiana Indians, and slightly higher than-roughly the same as Pima Indians).

    That is definitely interesting. I suspect it is for the same reason that height PGS don’t work properly for Pygmies (Pygmies are just the most extreme case we see). IIRC that is because some variants that are important for Pygmy height don’t show up in European GWAS. Amerindians are closer to Europeans than Africans genetically so I could see that effect existing between Amerindians and Africans there as well.

    I think this effect would result in Piffer’s PGS predicting higher IQs for Africa than the reality. If so, it is ironic that some possible issues Piffer’s critics highlight might result in the true numbers being different in a direction they don’t like.

    To tie this into the DUF1220 discussion elsewhere, this paper implicates DUF1220 in brain size: https://www.sciencedirect.com/science/article/pii/S0002929712003734
    They actually look at ethnic differences. They find no significant association, but given the sample sizes and p value I am suspicious.

    The ethnic distribution of individuals with known deletions or duplications was categorized as 26 white, 12 Hispanic, and 3 African American or other. Differences of copy ratio between ethnic groups were tested with a one-way ANOVA. We assessed population stratification in the group with a known deletion by including ethnicity as a covariate in a linear regression model of FOC Z score and CON1. There was no evidence of (1) a significant difference of copy ratio between ethnic groups (p > 0.10), (2) confounding by ethnicity, or (3) an association between ethnicity and FOC Z scores.

    Here is an old blog post looking at this: http://dienekes.blogspot.com/2012/08/our-big-human-brains-may-depend-on.html
    The cross species correlation is dramatic! And the human/Neandertal difference is very interesting.

    This paper does not cover DUF1220 (AFAICT), but gives an idea of the kind of CNV analysis which was possible in 2010: http://science.sciencemag.org/content/330/6004/641.full

    The author’s later paper looks at DUF1220 in the great ape lineage, but not intra-human differences: https://genome.cshlp.org/content/23/9/1373.full

    This 2015 paper by the same author looks at intra-human structural variations, but no mention of DUF1220 (interesting when dogs don’t bark): https://www.nature.com/articles/nature15394

    Figure S2B here: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5556342/bin/12864_2017_3976_MOESM2_ESM.pdf
    From this paper: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5556342/
    Shows boxplots by 1000 Genomes population for different areas of DUF1220 variation. The groups have somewhat differing means, but overlap heavily. Worth noting that the ASW population had high means. That figure looks like evidence to me that there are NOT systematic significant differences in DUF1220 copies by race.

    Worth repeating from the paper discussed in another comment: https://link.springer.com/article/10.1007%2Fs00439-014-1489-2

    Given these fndings, we examined associations between DUF1220 subtypes CON1 and CON2 and cognitive aptitude. We identifed a linear association between CON2 copy number and cognitive function in two independent populations of European descent.

    So CON2 is probably the region to focus on. Unfortunately there are relatively few copies there so Figure S2B is not really readable in that case.

    Read More
    • Replies: @Jm8
    Res


    Cont.

    "I suspect it is for the same reason that height PGS don’t work properly for Pygmies (Pygmies are just the most extreme case we see)..."

    True, they would be a more extreme case, but perhaps a significantly more (perhaps rather quite more) extreme case than other Africans who (except for Khoisans, who are even more divergent than Pygmies, and like Pygmies are descended from very divergent branches of modern humans), in most cases (other Africans groups that is—I believe) are actually somewhat closer to Eurasians than they are to Pygmies (because Eurasians and non-Pygmy/non-Khoisan Africans both derive from an East African branch of modern humans, which then split into Eurasians—the OOA—, and into those branches that stayed in Africa which became the majority of the ancestors of most African groups other than Pygmies/Khoisans.

    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  59. Jm8 says:

    Res

    “That is definitely interesting. I suspect it is for the same reason that height PGS don’t work properly for Pygmies (Pygmies are just the most extreme case we see). IIRC that is because some variants that are important for Pygmy height don’t show up in European GWAS. Amerindians are closer to Europeans than Africans genetically so I could see that effect existing between Amerindians and Africans there as well.

    I think this effect would result in Piffer’s PGS predicting higher IQs for Africa than the reality. If so, it is ironic that some possible issues Piffer’s critics highlight might result in the true numbers being different in a direction they don’t like.”

    It might, but it seems that it might also be that African numbers are (at least in some cases) are slightly underestimated; that genes for intelligence are not detected just as genes/variants reducing Pygmy height (unique to them, or found in/arising in mainly—or only—their population or more so than in others) are not detected in European GWAS. Or it could be as you say (that another effect is at play), and that true African numbers are slightly lower (at least in some groups), and perhaps (for instance) are in fact closer to Amerindian numbers.

    https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5556342/bin/12864_2017_3976_MOESM2_ESM.pdf
    From this paper: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5556342/
    Shows boxplots by 1000 Genomes population for different areas of DUF1220 variation. The groups have somewhat differing means, but overlap heavily. Worth noting that the ASW population had high means. That figure looks like evidence to me that there are NOT systematic significant differences in DUF1220 copies by race.”

    The DUF data is very interesting. I will have to continue looking over it.
    What does ASW stand for? Is it an African population (perhaps African South West)?

    Read More
    • Replies: @res

    It might, but it seems that it might also be
     
    Agreed that is possible. It just does not seem to be the way to bet given that it appears intelligence was positively selected for in the groups that left Africa.

    ASW is African ancestry in SW USA

    Here is a reference for all of the 1000 Genomes population abbreviations: https://en.wikipedia.org/wiki/1000_Genomes_Project#Human_genome_samples
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  60. Jm8 says:
    @res
    That is definitely interesting. I suspect it is for the same reason that height PGS don't work properly for Pygmies (Pygmies are just the most extreme case we see). IIRC that is because some variants that are important for Pygmy height don't show up in European GWAS. Amerindians are closer to Europeans than Africans genetically so I could see that effect existing between Amerindians and Africans there as well.

    I think this effect would result in Piffer's PGS predicting higher IQs for Africa than the reality. If so, it is ironic that some possible issues Piffer's critics highlight might result in the true numbers being different in a direction they don't like.

    To tie this into the DUF1220 discussion elsewhere, this paper implicates DUF1220 in brain size: https://www.sciencedirect.com/science/article/pii/S0002929712003734
    They actually look at ethnic differences. They find no significant association, but given the sample sizes and p value I am suspicious.

    The ethnic distribution of individuals with known deletions or duplications was categorized as 26 white, 12 Hispanic, and 3 African American or other. Differences of copy ratio between ethnic groups were tested with a one-way ANOVA. We assessed population stratification in the group with a known deletion by including ethnicity as a covariate in a linear regression model of FOC Z score and CON1. There was no evidence of (1) a significant difference of copy ratio between ethnic groups (p > 0.10), (2) confounding by ethnicity, or (3) an association between ethnicity and FOC Z scores.
     
    Here is an old blog post looking at this: http://dienekes.blogspot.com/2012/08/our-big-human-brains-may-depend-on.html
    The cross species correlation is dramatic! And the human/Neandertal difference is very interesting.

    http://2.bp.blogspot.com/-tTV5L6PVMg4/UC0-6nXIKvI/AAAAAAAAFa8/dzu_4yBjwR8/s320/brain.jpg

    This paper does not cover DUF1220 (AFAICT), but gives an idea of the kind of CNV analysis which was possible in 2010: http://science.sciencemag.org/content/330/6004/641.full

    https://d2ufo47lrtsv5s.cloudfront.net/content/sci/330/6004/641/F3.large.jpg

    The author's later paper looks at DUF1220 in the great ape lineage, but not intra-human differences: https://genome.cshlp.org/content/23/9/1373.full

    This 2015 paper by the same author looks at intra-human structural variations, but no mention of DUF1220 (interesting when dogs don't bark): https://www.nature.com/articles/nature15394

    Figure S2B here: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5556342/bin/12864_2017_3976_MOESM2_ESM.pdf
    From this paper: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5556342/
    Shows boxplots by 1000 Genomes population for different areas of DUF1220 variation. The groups have somewhat differing means, but overlap heavily. Worth noting that the ASW population had high means. That figure looks like evidence to me that there are NOT systematic significant differences in DUF1220 copies by race.

    Worth repeating from the paper discussed in another comment: https://link.springer.com/article/10.1007%2Fs00439-014-1489-2

    Given these fndings, we examined associations between DUF1220 subtypes CON1 and CON2 and cognitive aptitude. We identifed a linear association between CON2 copy number and cognitive function in two independent populations of European descent.

     

    So CON2 is probably the region to focus on. Unfortunately there are relatively few copies there so Figure S2B is not really readable in that case.

    Res

    Cont.

    “I suspect it is for the same reason that height PGS don’t work properly for Pygmies (Pygmies are just the most extreme case we see)…”

    True, they would be a more extreme case, but perhaps a significantly more (perhaps rather quite more) extreme case than other Africans who (except for Khoisans, who are even more divergent than Pygmies, and like Pygmies are descended from very divergent branches of modern humans), in most cases (other Africans groups that is—I believe) are actually somewhat closer to Eurasians than they are to Pygmies (because Eurasians and non-Pygmy/non-Khoisan Africans both derive from an East African branch of modern humans, which then split into Eurasians—the OOA—, and into those branches that stayed in Africa which became the majority of the ancestors of most African groups other than Pygmies/Khoisans.

    Read More
    • Replies: @Jm8
    Edit:

    "True, they would be a more extreme case, but perhaps a significantly more (perhaps rather quite more) extreme case than other Africans who (except for Khoisans, who are even more divergent than Pygmies, " "...though of course other (non-Khoisan/non-Pygmy) Africans are also quite divergent relative to Eurasians and would likely have variants not present in Europeans (and other Eurasians)."
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  61. res says:
    @Jm8
    Res

    "That is definitely interesting. I suspect it is for the same reason that height PGS don’t work properly for Pygmies (Pygmies are just the most extreme case we see). IIRC that is because some variants that are important for Pygmy height don’t show up in European GWAS. Amerindians are closer to Europeans than Africans genetically so I could see that effect existing between Amerindians and Africans there as well.

    I think this effect would result in Piffer’s PGS predicting higher IQs for Africa than the reality. If so, it is ironic that some possible issues Piffer’s critics highlight might result in the true numbers being different in a direction they don’t like."


    It might, but it seems that it might also be that African numbers are (at least in some cases) are slightly underestimated; that genes for intelligence are not detected just as genes/variants reducing Pygmy height (unique to them, or found in/arising in mainly—or only—their population or more so than in others) are not detected in European GWAS. Or it could be as you say (that another effect is at play), and that true African numbers are slightly lower (at least in some groups), and perhaps (for instance) are in fact closer to Amerindian numbers.



    "https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5556342/bin/12864_2017_3976_MOESM2_ESM.pdf
    From this paper: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5556342/
    Shows boxplots by 1000 Genomes population for different areas of DUF1220 variation. The groups have somewhat differing means, but overlap heavily. Worth noting that the ASW population had high means. That figure looks like evidence to me that there are NOT systematic significant differences in DUF1220 copies by race."

    The DUF data is very interesting. I will have to continue looking over it.
    What does ASW stand for? Is it an African population (perhaps African South West)?

    It might, but it seems that it might also be

    Agreed that is possible. It just does not seem to be the way to bet given that it appears intelligence was positively selected for in the groups that left Africa.

    ASW is African ancestry in SW USA

    Here is a reference for all of the 1000 Genomes population abbreviations: https://en.wikipedia.org/wiki/1000_Genomes_Project#Human_genome_samples

    Read More
    • Replies: @Jm8
    "Agreed that is possible. It just does not seem to be the way to bet given that it appears intelligence was positively selected for in the groups that left Africa."

    I doubt (or am skeptical) that this is true of non-Africans vs. Africans across the board.
    I don't meant suggest (for example) that the true numbers for Africans would be higher than those of Europeans or East Asians—unlikely (or anything like that). But it would be plausible that they could be roughly equal to (and maybe even a bit higher in some/certain cases) than those of Amerindians. It is not implausible that the true numbers for Mandinka or Yoruba for example (iron age/iron using groups that created kingdoms, empires and city states)—and even some Bantu—could be higher than those of the (traditionally) neolithic small-scale horticulturalist Karitiana or Surui (for instance), or roughly equal to groups such as the Pima or even Maya.


    Also, the San are more genetically divergent than Pygmies, but get lower scores than them, which is interesting. I also suspect that the extremely low scores of Oceanians may not be quite representative of "true numbers" (but we may know more with more data in the future).

    , @Factorize
    res, I have been modifying the R code to find the chromosomal level numbers for the groups given. Did you notice that there only seems to be about 100 of the SNPs in the dataset? Why weren't the 1000 Genomes genotypes used? When I downloaded from Genome Browser they had almost all 3,000 of the SNPs for 2500 people. Looks like it should be possible to modify the R code to wind up with individual level PGS by chromosomal strand.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  62. Jm8 says:
    @Jm8
    Res


    Cont.

    "I suspect it is for the same reason that height PGS don’t work properly for Pygmies (Pygmies are just the most extreme case we see)..."

    True, they would be a more extreme case, but perhaps a significantly more (perhaps rather quite more) extreme case than other Africans who (except for Khoisans, who are even more divergent than Pygmies, and like Pygmies are descended from very divergent branches of modern humans), in most cases (other Africans groups that is—I believe) are actually somewhat closer to Eurasians than they are to Pygmies (because Eurasians and non-Pygmy/non-Khoisan Africans both derive from an East African branch of modern humans, which then split into Eurasians—the OOA—, and into those branches that stayed in Africa which became the majority of the ancestors of most African groups other than Pygmies/Khoisans.

    Edit:

    “True, they would be a more extreme case, but perhaps a significantly more (perhaps rather quite more) extreme case than other Africans who (except for Khoisans, who are even more divergent than Pygmies, ” “…though of course other (non-Khoisan/non-Pygmy) Africans are also quite divergent relative to Eurasians and would likely have variants not present in Europeans (and other Eurasians).”

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  63. Jm8 says:
    @res

    It might, but it seems that it might also be
     
    Agreed that is possible. It just does not seem to be the way to bet given that it appears intelligence was positively selected for in the groups that left Africa.

    ASW is African ancestry in SW USA

    Here is a reference for all of the 1000 Genomes population abbreviations: https://en.wikipedia.org/wiki/1000_Genomes_Project#Human_genome_samples

    “Agreed that is possible. It just does not seem to be the way to bet given that it appears intelligence was positively selected for in the groups that left Africa.”

    I doubt (or am skeptical) that this is true of non-Africans vs. Africans across the board.
    I don’t meant suggest (for example) that the true numbers for Africans would be higher than those of Europeans or East Asians—unlikely (or anything like that). But it would be plausible that they could be roughly equal to (and maybe even a bit higher in some/certain cases) than those of Amerindians. It is not implausible that the true numbers for Mandinka or Yoruba for example (iron age/iron using groups that created kingdoms, empires and city states)—and even some Bantu—could be higher than those of the (traditionally) neolithic small-scale horticulturalist Karitiana or Surui (for instance), or roughly equal to groups such as the Pima or even Maya.

    Also, the San are more genetically divergent than Pygmies, but get lower scores than them, which is interesting. I also suspect that the extremely low scores of Oceanians may not be quite representative of “true numbers” (but we may know more with more data in the future).

    Read More
    • Replies: @Jm8
    Edit:

    "“Agreed that is possible. It just does not seem to be the way to bet given that it appears intelligence was positively selected for in the groups that left Africa.”"

    "I doubt (or am skeptical) that this is true of non-Africans vs. Africans across the board (and unsure that is true at all/generally, though of course it is possible)."

    , @Johan Meyer
    Most sub-Saharan African ethnic groups worked iron. Few worked bronze. A more fruitful study would be to check the connection between bronze work and polygenic IQ results.

    https://archive.org/details/in.ernet.dli.2015.57373
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  64. Jm8 says:
    @Jm8
    "Agreed that is possible. It just does not seem to be the way to bet given that it appears intelligence was positively selected for in the groups that left Africa."

    I doubt (or am skeptical) that this is true of non-Africans vs. Africans across the board.
    I don't meant suggest (for example) that the true numbers for Africans would be higher than those of Europeans or East Asians—unlikely (or anything like that). But it would be plausible that they could be roughly equal to (and maybe even a bit higher in some/certain cases) than those of Amerindians. It is not implausible that the true numbers for Mandinka or Yoruba for example (iron age/iron using groups that created kingdoms, empires and city states)—and even some Bantu—could be higher than those of the (traditionally) neolithic small-scale horticulturalist Karitiana or Surui (for instance), or roughly equal to groups such as the Pima or even Maya.


    Also, the San are more genetically divergent than Pygmies, but get lower scores than them, which is interesting. I also suspect that the extremely low scores of Oceanians may not be quite representative of "true numbers" (but we may know more with more data in the future).

    Edit:

    ““Agreed that is possible. It just does not seem to be the way to bet given that it appears intelligence was positively selected for in the groups that left Africa.””

    “I doubt (or am skeptical) that this is true of non-Africans vs. Africans across the board (and unsure that is true at all/generally, though of course it is possible).”

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

    It might, but it seems that it might also be
     
    Agreed that is possible. It just does not seem to be the way to bet given that it appears intelligence was positively selected for in the groups that left Africa.

    ASW is African ancestry in SW USA

    Here is a reference for all of the 1000 Genomes population abbreviations: https://en.wikipedia.org/wiki/1000_Genomes_Project#Human_genome_samples

    res, I have been modifying the R code to find the chromosomal level numbers for the groups given. Did you notice that there only seems to be about 100 of the SNPs in the dataset? Why weren’t the 1000 Genomes genotypes used? When I downloaded from Genome Browser they had almost all 3,000 of the SNPs for 2500 people. Looks like it should be possible to modify the R code to wind up with individual level PGS by chromosomal strand.

    Read More
    • Replies: @res
    Do you mean the Racimo data set I talked about in comment 37? Is your <100 the same as my 78 in comment 40? If so I think the issue is as I mentioned there and they trimmed SNPs from their neutral file. Not sure if they just trimmed all the phenotypes they were testing or if they had other criteria. In theory you could pull the missing SNPs from 1000 Genomes. Do you have any idea how hard it would be to extract the ancestral/derived allele frequency data they use for each SNP/population?
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  66. res says:
    @Factorize
    res, I have been modifying the R code to find the chromosomal level numbers for the groups given. Did you notice that there only seems to be about 100 of the SNPs in the dataset? Why weren't the 1000 Genomes genotypes used? When I downloaded from Genome Browser they had almost all 3,000 of the SNPs for 2500 people. Looks like it should be possible to modify the R code to wind up with individual level PGS by chromosomal strand.

    Do you mean the Racimo data set I talked about in comment 37? Is your <100 the same as my 78 in comment 40? If so I think the issue is as I mentioned there and they trimmed SNPs from their neutral file. Not sure if they just trimmed all the phenotypes they were testing or if they had other criteria. In theory you could pull the missing SNPs from 1000 Genomes. Do you have any idea how hard it would be to extract the ancestral/derived allele frequency data they use for each SNP/population?

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

    res, the results_10k_final file had 6731 input lines with 127 sets of 53.
    They only used 127 SNPs? I would like someone to try it out using the individual genotypes from the 1000 Genomes. I suppose that there will be a fair amount of research response to this so we can just sit back. Has the UKBB been phenotyped for EA? (Not sure whether this current article is using the UKBB sample.) It would be very helpful to have a phenotyped sample instead of simply assuming that the sample is representative of average phenotypes in the population. Do you have any idea when the Nature Genetics article might go online? It was a great idea to have this out there and building up some excitement beforehand, though it would be nice now to actually finally see the article. I will be very interested to see whether this attracts any mainstream interest. How couldn’t it? This research could fundamentally change human reality.

    Read More
    • Replies: @res
    Racimo et al. used 1000 Genomes and another dataset. See their methods and materials section pp. 4-5:

    We fitted admixture graphs using genome-wide data from Phase 3 of the 1000 Genomes Project (Auton et al. 2015) and a more broadly sampled SNP chip dataset of present-day humans from populations genotyped with the Human Origins array (Patterson et al. 2012; Lazaridis et al. 2014). The latter dataset was imputed using SHAPEIT (Delaneau et al. 2013) on the Michigan Imputation Server (Das et al. 2016) with the 1000 Genomes data as the reference panel (Bhérer et al. in prep.).
     
    Regarding the UKBB.

    There appear to be multiple types of phenotypic data. Here are the biomarkers they offer: http://www.ukbiobank.ac.uk/wp-content/uploads/2013/11/BCM023_ukb_biomarker_panel_website_v1.0-Aug-2015.pdf
    And a timeline for future data availability: http://biobank.ctsu.ox.ac.uk/crystal/exinfo.cgi?src=future_timelines

    Here are all of the data fields by type. It looks like they did a comprehensive survey (similar to US NHANES) in addition to the genotyping: http://biobank.ctsu.ox.ac.uk/crystal/list.cgi

    This page is probably the best way to browse their data to see what it contains: http://biobank.ctsu.ox.ac.uk/crystal/label.cgi
    For example, notice the UK Biobank Assessment Centre > Cognitive function section:

    Reaction time
    Numeric memory
    Fluid intelligence
    Trail making
    Matrix pattern completion
    Tower rearranging
    Picture vocabulary
    Symbol digit substitution
    Paired associate learning
    Prospective memory
    Pairs matching
    Lights pattern memory
    Word production

     

    Not all participants are tested for these. For example, fluid intelligence has about 190k participants and covers about a +- 3SD range: http://biobank.ctsu.ox.ac.uk/crystal/field.cgi?id=20016

    From that page you can navigate to summary data for each field (including # participants by field and summary statistics!). For example, here are the two fields which look most relevant for EA (i.e. yes, the UKBB has EA information): http://biobank.ctsu.ox.ac.uk/crystal/field.cgi?id=6138 (best coverage-~500k participants-so probably what everyone is using)
    http://biobank.ctsu.ox.ac.uk/crystal/field.cgi?id=845
    Their SNP data has about 488k participants: http://biobank.ctsu.ox.ac.uk/crystal/field.cgi?id=22101
    They also offer a search capability: http://biobank.ctsu.ox.ac.uk/crystal/search.cgi

    One thing that confused me is for height. I see http://biobank.ctsu.ox.ac.uk/crystal/field.cgi?id=12144 which only has 24k participants covering a +-3.5SD range. This explains it. From the notes: "Height measured prior to imaging stages. Required by DXA device for calibration."
    Here is the real "Standing Height" measurement: http://biobank.ctsu.ox.ac.uk/crystal/field.cgi?id=50
    544k participants covering a range from -10 (!) SD to 4.4 SD.
    That 75cm height is shocking given the age range is 37-73: http://biobank.ctsu.ox.ac.uk/crystal/field.cgi?id=21022
    Their distributionplot only goes down to 133cm.

    The UKBB data browser is great IMHO.

    Here is a data dictionary type file describing the UKBB genetic data. I did not realize they included HLA imputation. There also appears to be some support for CNVs, though not actual copy numbers (?). And here is another description: http://www.ukbiobank.ac.uk/scientists-3/genetic-data/
    http://www.ukbiobank.ac.uk/wp-content/uploads/2017/07/ukb_genetic_file_description.txt

    P.S. I have no inside information on the Lee article publication date. I'm just waiting for either Dr. Thompson or Steve Hsu to post about it ; )
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  68. res says:
    @Factorize
    res, the results_10k_final file had 6731 input lines with 127 sets of 53.
    They only used 127 SNPs? I would like someone to try it out using the individual genotypes from the 1000 Genomes. I suppose that there will be a fair amount of research response to this so we can just sit back. Has the UKBB been phenotyped for EA? (Not sure whether this current article is using the UKBB sample.) It would be very helpful to have a phenotyped sample instead of simply assuming that the sample is representative of average phenotypes in the population. Do you have any idea when the Nature Genetics article might go online? It was a great idea to have this out there and building up some excitement beforehand, though it would be nice now to actually finally see the article. I will be very interested to see whether this attracts any mainstream interest. How couldn't it? This research could fundamentally change human reality.

    Racimo et al. used 1000 Genomes and another dataset. See their methods and materials section pp. 4-5:

    We fitted admixture graphs using genome-wide data from Phase 3 of the 1000 Genomes Project (Auton et al. 2015) and a more broadly sampled SNP chip dataset of present-day humans from populations genotyped with the Human Origins array (Patterson et al. 2012; Lazaridis et al. 2014). The latter dataset was imputed using SHAPEIT (Delaneau et al. 2013) on the Michigan Imputation Server (Das et al. 2016) with the 1000 Genomes data as the reference panel (Bhérer et al. in prep.).

    Regarding the UKBB.

    There appear to be multiple types of phenotypic data. Here are the biomarkers they offer: http://www.ukbiobank.ac.uk/wp-content/uploads/2013/11/BCM023_ukb_biomarker_panel_website_v1.0-Aug-2015.pdf
    And a timeline for future data availability: http://biobank.ctsu.ox.ac.uk/crystal/exinfo.cgi?src=future_timelines

    Here are all of the data fields by type. It looks like they did a comprehensive survey (similar to US NHANES) in addition to the genotyping: http://biobank.ctsu.ox.ac.uk/crystal/list.cgi

    This page is probably the best way to browse their data to see what it contains: http://biobank.ctsu.ox.ac.uk/crystal/label.cgi
    For example, notice the UK Biobank Assessment Centre > Cognitive function section:

    Reaction time
    Numeric memory
    Fluid intelligence
    Trail making
    Matrix pattern completion
    Tower rearranging
    Picture vocabulary
    Symbol digit substitution
    Paired associate learning
    Prospective memory
    Pairs matching
    Lights pattern memory
    Word production

    Not all participants are tested for these. For example, fluid intelligence has about 190k participants and covers about a +- 3SD range: http://biobank.ctsu.ox.ac.uk/crystal/field.cgi?id=20016

    From that page you can navigate to summary data for each field (including # participants by field and summary statistics!). For example, here are the two fields which look most relevant for EA (i.e. yes, the UKBB has EA information): http://biobank.ctsu.ox.ac.uk/crystal/field.cgi?id=6138 (best coverage-~500k participants-so probably what everyone is using)

    http://biobank.ctsu.ox.ac.uk/crystal/field.cgi?id=845

    Their SNP data has about 488k participants: http://biobank.ctsu.ox.ac.uk/crystal/field.cgi?id=22101
    They also offer a search capability: http://biobank.ctsu.ox.ac.uk/crystal/search.cgi

    One thing that confused me is for height. I see http://biobank.ctsu.ox.ac.uk/crystal/field.cgi?id=12144 which only has 24k participants covering a +-3.5SD range. This explains it. From the notes: “Height measured prior to imaging stages. Required by DXA device for calibration.”
    Here is the real “Standing Height” measurement: http://biobank.ctsu.ox.ac.uk/crystal/field.cgi?id=50
    544k participants covering a range from -10 (!) SD to 4.4 SD.
    That 75cm height is shocking given the age range is 37-73: http://biobank.ctsu.ox.ac.uk/crystal/field.cgi?id=21022
    Their distributionplot only goes down to 133cm.

    The UKBB data browser is great IMHO.

    Here is a data dictionary type file describing the UKBB genetic data. I did not realize they included HLA imputation. There also appears to be some support for CNVs, though not actual copy numbers (?). And here is another description: http://www.ukbiobank.ac.uk/scientists-3/genetic-data/

    http://www.ukbiobank.ac.uk/wp-content/uploads/2017/07/ukb_genetic_file_description.txt

    P.S. I have no inside information on the Lee article publication date. I’m just waiting for either Dr. Thompson or Steve Hsu to post about it ; )

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

    Thank you res!

    Startling how many phenotypes they appear to have acquired.
    If you click on Index on top menu, Catalogues, Fields, and Categorical(single), there appears to be hundreds of phenotypes! The other choices after Fields also have numerous choices. There might be over 1,000 measured phenotypes. Amazing! Notice the last item in the Categorical(single) list: Yorkshire pudding intake. Are they kidding?

    The UKBB was able to put so many of the pieces together to wind up with this overwhelming success. Collecting a large number of phenotypes was one part of the success, as was hitting an appropriate scale so that useful results could be found with the dataset, as was making sure that the cost structure did not become so bloated that ultimately the government would figure that it was fiscally unjustified. When I looked through their financials it was surprising how minimal the expenses of the shell company actually were; they managed to offload nearly all the costs to external users etc. . At the same time nearly all the investment and publications flow back to UKBB and build the brand. The way it is structured and given its impressive level of performance, it is almost impossible to defund it.

    While so many of the other BB attempts have been notable disappointments; a few have went bankrupt, others seem so late to the game it is difficult to know whether they will be able to make any meaningful contribution. With other BBs, they seem to have the data, though for whatever reason they have not published it {I am thinking here of EPIC with their Educational Attainment dataset.}. For me this is a great example of why endlessly consolidating governments into yet larger entities often is such a big mistake. Without some element of competition or at least separately managed projects, it might never be obvious that there could be a range of achievement levels possible. The main distinguishing feature of success with UKBB has not been so much related to the science as to the management.

    res, you notice that UKBB is going full genome! Yeah! I was wondering about that for the last while.
    Could they find someone to provide the dime for the sequencing? Yes, looks like they want to start with 50k and then move up to 500k! This is just awesome! They will have full exomes ready by end of next year, and full genomes will ramp up over the next year or two. The 500k full genomes will
    cost about 300 M pounds at current prices, though it would not be unlikely that they could wind up costing quite bit less than that on the final bill. So we’re down to 600 pounds for a full sequence.
    It seems that around 100 the entire consumer marketplace would unlock and we would have tens of millions of full genome sequences.

    http://www.ukbiobank.ac.uk/2018/04/whole-genome-sequencing-will-transform-the-research-landscape-for-a-wide-range-of-diseases/

    About the only question mark for me is why they haven’t expanded the sample size. Considering the benefits involved and the fairly modest costs (which tend to become even more modest through time), topping up the sample at least for a phase II to 1M does not seem unreasonable. Perhaps at some point they might even be able to include participants who already had their genomes sequenced. UKBB might then do a quality control run to make ensure the accuracy of the submitted sequence. Possibly a 5x full genome confirmation with a gene chip.

    Also wondering whether they will use nanopores with the full genome sequences. Possibly could be somewhat more expensive, though a 3-5x coverage full genome with the super long reads from nanopores would give you a nice phased genome. This would allow verification of the phasing that has been done on the dataset. Also will be interested to see what other sequencing they might feel will be worthwhile (e.g. methylome etc.).

    Having full genomes for 500K could give us extraordinary into the genetics of height etc. We have almost maxed out using current additive approaches with available SNPs. It will clearly be fascinating to see how much more the full genomes will give. The press release notes that the scientists were jubilant about the full genomes announcement. I’ll bet they were! This might truly unlock the human genome.

    Read More
    • Replies: @res

    res, you notice that UKBB is going full genome!
     
    I saw something about that, but completion is a ways off, right? I wonder if they are planning on making CNV data available either now or in the future.

    The UKBB really is an incredible resource which is in the process of revolutionizing genomics.

    The Million Veteran Program (MVP) also has potential in this area. Emil mentioned it earlier in this thread and my later comment gives some links: http://www.unz.com/jthompson/the-hsu-boundary/#comment-2007892
    The AFQT data they have is probably as good a measure of cognitive ability as we will see in large sample genomics studies (I would love to hear about better examples!).
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  70. res says:
    @Factorize
    Thank you res!

    Startling how many phenotypes they appear to have acquired.
    If you click on Index on top menu, Catalogues, Fields, and Categorical(single), there appears to be hundreds of phenotypes! The other choices after Fields also have numerous choices. There might be over 1,000 measured phenotypes. Amazing! Notice the last item in the Categorical(single) list: Yorkshire pudding intake. Are they kidding?

    The UKBB was able to put so many of the pieces together to wind up with this overwhelming success. Collecting a large number of phenotypes was one part of the success, as was hitting an appropriate scale so that useful results could be found with the dataset, as was making sure that the cost structure did not become so bloated that ultimately the government would figure that it was fiscally unjustified. When I looked through their financials it was surprising how minimal the expenses of the shell company actually were; they managed to offload nearly all the costs to external users etc. . At the same time nearly all the investment and publications flow back to UKBB and build the brand. The way it is structured and given its impressive level of performance, it is almost impossible to defund it.

    While so many of the other BB attempts have been notable disappointments; a few have went bankrupt, others seem so late to the game it is difficult to know whether they will be able to make any meaningful contribution. With other BBs, they seem to have the data, though for whatever reason they have not published it {I am thinking here of EPIC with their Educational Attainment dataset.}. For me this is a great example of why endlessly consolidating governments into yet larger entities often is such a big mistake. Without some element of competition or at least separately managed projects, it might never be obvious that there could be a range of achievement levels possible. The main distinguishing feature of success with UKBB has not been so much related to the science as to the management.

    res, you notice that UKBB is going full genome! Yeah! I was wondering about that for the last while.
    Could they find someone to provide the dime for the sequencing? Yes, looks like they want to start with 50k and then move up to 500k! This is just awesome! They will have full exomes ready by end of next year, and full genomes will ramp up over the next year or two. The 500k full genomes will
    cost about 300 M pounds at current prices, though it would not be unlikely that they could wind up costing quite bit less than that on the final bill. So we're down to 600 pounds for a full sequence.
    It seems that around 100 the entire consumer marketplace would unlock and we would have tens of millions of full genome sequences.

    http://www.ukbiobank.ac.uk/2018/04/whole-genome-sequencing-will-transform-the-research-landscape-for-a-wide-range-of-diseases/

    About the only question mark for me is why they haven't expanded the sample size. Considering the benefits involved and the fairly modest costs (which tend to become even more modest through time), topping up the sample at least for a phase II to 1M does not seem unreasonable. Perhaps at some point they might even be able to include participants who already had their genomes sequenced. UKBB might then do a quality control run to make ensure the accuracy of the submitted sequence. Possibly a 5x full genome confirmation with a gene chip.

    Also wondering whether they will use nanopores with the full genome sequences. Possibly could be somewhat more expensive, though a 3-5x coverage full genome with the super long reads from nanopores would give you a nice phased genome. This would allow verification of the phasing that has been done on the dataset. Also will be interested to see what other sequencing they might feel will be worthwhile (e.g. methylome etc.).

    Having full genomes for 500K could give us extraordinary into the genetics of height etc. We have almost maxed out using current additive approaches with available SNPs. It will clearly be fascinating to see how much more the full genomes will give. The press release notes that the scientists were jubilant about the full genomes announcement. I'll bet they were! This might truly unlock the human genome.

    res, you notice that UKBB is going full genome!

    I saw something about that, but completion is a ways off, right? I wonder if they are planning on making CNV data available either now or in the future.

    The UKBB really is an incredible resource which is in the process of revolutionizing genomics.

    The Million Veteran Program (MVP) also has potential in this area. Emil mentioned it earlier in this thread and my later comment gives some links: http://www.unz.com/jthompson/the-hsu-boundary/#comment-2007892
    The AFQT data they have is probably as good a measure of cognitive ability as we will see in large sample genomics studies (I would love to hear about better examples!).

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  71. @Jm8
    "Agreed that is possible. It just does not seem to be the way to bet given that it appears intelligence was positively selected for in the groups that left Africa."

    I doubt (or am skeptical) that this is true of non-Africans vs. Africans across the board.
    I don't meant suggest (for example) that the true numbers for Africans would be higher than those of Europeans or East Asians—unlikely (or anything like that). But it would be plausible that they could be roughly equal to (and maybe even a bit higher in some/certain cases) than those of Amerindians. It is not implausible that the true numbers for Mandinka or Yoruba for example (iron age/iron using groups that created kingdoms, empires and city states)—and even some Bantu—could be higher than those of the (traditionally) neolithic small-scale horticulturalist Karitiana or Surui (for instance), or roughly equal to groups such as the Pima or even Maya.


    Also, the San are more genetically divergent than Pygmies, but get lower scores than them, which is interesting. I also suspect that the extremely low scores of Oceanians may not be quite representative of "true numbers" (but we may know more with more data in the future).

    Most sub-Saharan African ethnic groups worked iron. Few worked bronze. A more fruitful study would be to check the connection between bronze work and polygenic IQ results.

    https://archive.org/details/in.ernet.dli.2015.57373

    Read More
    • Replies: @Jm8
    Iron working (and smelting) is not an easy thing (it's quite complex and difficult, and requiring high temperatures among other things—and further innovations like carbon steel were also invented among a few cultures e.g. the Haya tribe of West Tanzania) (and is evident early in S.E. and Nigeria about as Early as in Eurasia), and I think there is no reason to associate bronze more with intelligence.

    But anyway, bronze and brass working (as well as copper working), along with iron, were common all over West Africa and some parts of Central Africa (Including—in West Africa, where bronze and brass are common: the many Yoruba kingdoms: Ife, Owe, Ijebu, Owe, Oyo and other subgroups etc); the Igbo—such as at Igbo Ukwu—and other South East Nigerian groups—who produced several styles of bronze work over the centuries incl. that called the "Lower Niger Bronze Industry; the Edo of the Benin kingdom and related tribes; as well as those of Central and Northern Nigeria; the Ashanti/Akan—including for gold weights, sculptures, etc—, Ewe, Ga and other ethnic groups in Ghana—both North and South; the peoples of Dahomey; cultures of the Senegambia and Mali such as among many Mande speaking tribes—like the Soninke, Mandinka, Bambara etc‚—including those of Djenne/Jenne-Jeno and other city states along the Niger—there are several early-mid medieval bronzes surviving from there; the many tribes of the Voltaic regions in Burkina and Niger, and other areas. In Central Africa bronze, brass and copper were sometimes worked (though in Central Africa, copper was the more common) by cultures like the Bakongo (with their various subgroups), the Bakuba, the Baluba and Songye, and various other tribes in the Central Africa region (and some in the East Central African lakes region in the Uganda/Rwanda area. Some tribes in Eastern Africa (such as Tanzania also worked some of those metals. Bronze was also traditionally worked by cultures in Cameroon (like the kingdoms of the Bamileke, Bamoum, Bafut, etc, and tribes in Liberia/Sierra Leone. I know less about southern Africa (and bronze and brass may have been rarer in southern Africa).

    An overview—first link below (albeit somewhat brief, but inclusive) of the occurrence of bronze working from various bronze working cultures (with a focus on West Africa—where a great many tribes practiced it over large areas).

    http://www.sahistory.org.za/sites/default/files/DC/asjan59.17/asjan59.17.pdf


    A few examples:

    https://www.metmuseum.org/art/collection/search/310770
    https://www.metmuseum.org/art/collection/search/310257

    https://www.google.com/search?q=Ife+bronze&source=lnms&tbm=isch&sa=X&ved=0ahUKEwj4oL227u_aAhVxplkKHW9fBnoQ_AUICigB&biw=1015&bih=784

    Bamoum Bronze from Cameroon:

    gong at left link below w/ sculpted spider
    https://www.pinterest.com/pin/473933560769434257/

    https://auction.catawiki.com/kavels/9045487-large-bronze-pipe-bamun-bamileke-cameroon

    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  72. Jm8 says:
    @Johan Meyer
    Most sub-Saharan African ethnic groups worked iron. Few worked bronze. A more fruitful study would be to check the connection between bronze work and polygenic IQ results.

    https://archive.org/details/in.ernet.dli.2015.57373

    Iron working (and smelting) is not an easy thing (it’s quite complex and difficult, and requiring high temperatures among other things—and further innovations like carbon steel were also invented among a few cultures e.g. the Haya tribe of West Tanzania) (and is evident early in S.E. and Nigeria about as Early as in Eurasia), and I think there is no reason to associate bronze more with intelligence.

    But anyway, bronze and brass working (as well as copper working), along with iron, were common all over West Africa and some parts of Central Africa (Including—in West Africa, where bronze and brass are common: the many Yoruba kingdoms: Ife, Owe, Ijebu, Owe, Oyo and other subgroups etc); the Igbo—such as at Igbo Ukwu—and other South East Nigerian groups—who produced several styles of bronze work over the centuries incl. that called the “Lower Niger Bronze Industry; the Edo of the Benin kingdom and related tribes; as well as those of Central and Northern Nigeria; the Ashanti/Akan—including for gold weights, sculptures, etc—, Ewe, Ga and other ethnic groups in Ghana—both North and South; the peoples of Dahomey; cultures of the Senegambia and Mali such as among many Mande speaking tribes—like the Soninke, Mandinka, Bambara etc‚—including those of Djenne/Jenne-Jeno and other city states along the Niger—there are several early-mid medieval bronzes surviving from there; the many tribes of the Voltaic regions in Burkina and Niger, and other areas. In Central Africa bronze, brass and copper were sometimes worked (though in Central Africa, copper was the more common) by cultures like the Bakongo (with their various subgroups), the Bakuba, the Baluba and Songye, and various other tribes in the Central Africa region (and some in the East Central African lakes region in the Uganda/Rwanda area. Some tribes in Eastern Africa (such as Tanzania also worked some of those metals. Bronze was also traditionally worked by cultures in Cameroon (like the kingdoms of the Bamileke, Bamoum, Bafut, etc, and tribes in Liberia/Sierra Leone. I know less about southern Africa (and bronze and brass may have been rarer in southern Africa).

    An overview—first link below (albeit somewhat brief, but inclusive) of the occurrence of bronze working from various bronze working cultures (with a focus on West Africa—where a great many tribes practiced it over large areas).

    http://www.sahistory.org.za/sites/default/files/DC/asjan59.17/asjan59.17.pdf

    A few examples:

    https://www.metmuseum.org/art/collection/search/310770

    https://www.metmuseum.org/art/collection/search/310257

    https://www.google.com/search?q=Ife+bronze&source=lnms&tbm=isch&sa=X&ved=0ahUKEwj4oL227u_aAhVxplkKHW9fBnoQ_AUICigB&biw=1015&bih=784

    Bamoum Bronze from Cameroon:

    gong at left link below w/ sculpted spider

    https://www.pinterest.com/pin/473933560769434257/

    https://auction.catawiki.com/kavels/9045487-large-bronze-pipe-bamun-bamileke-cameroon

    Read More
    • Replies: @Jm8
    Edit: "...cultures of the Senegambia and Mali such as among many Mande speaking tribes—like the Soninke, Mandinka, Bambara etc‚—including those of Djenne/Jenne-Jeno and other city states along the Niger river—there are many early-mid medieval bronze sculptures, jewelry, and weapons/tools surviving from there"


    from Djenne:

    (1st link—should be image result at top toward right: abstract ornament with stylized pointing hands, outward hooks on either side, and small human figure in between)

    https://www.google.com/search?biw=1138&bih=885&tbm=isch&sa=1&ei=aVnuWuzoG-e1ggfn3ragDg&q=djenne+human+bronze+ornament&oq=djenne+human+bronze+ornament&gs_l=img.3...179415.180081.0.180277.6.6.0.0.0.0.77.363.6.6.0....0...1c.1.64.img..0.0.0....0.WmVGhibkNHc#imgrc=cioRZjUKhG_TKM:


    https://www.metmuseum.org/art/collection/search/310001

    https://www.google.com/search?biw=1089&bih=794&tbm=isch&sa=1&ei=71buWtuYDoXk_Abiw6GABA&q=djenne+bronze+&oq=djenne+bronze+&gs_l=img.3...270436.270436.0.270555.1.1.0.0.0.0.66.66.1.1.0....0...1c.1.64.img..0.0.0....0.84DAQiOp28Q#imgrc=rHijkUQ75jas5M:
    , @Johan Meyer
    Some comments:

    If genes adding to intelligence enable bronze-work, then certainly the presence of the genes would have to precede the metal work. If, however, the genes are a response to an intelligence-destroying poison in the source material (soft metal deposits usually contain lead), then bronze work should precede the elevated presence of the genes. In this regard, genetically variable sensitivity to environment violates the assumption of separate (additive with separate terms) genetic and environmental contribution to intelligence, in which case it becomes meaningless to speak of x percent contribution of environment (or genes) to intelligence, as genes are implicated in the effect of environment in such models (dose response as a function of genes).

    Most of the objects are art, status objects or ritual objects. Tools or weapons are rare. You have not addressed the matter of wide-spread iron metallurgy, especially regarding the link that I posted. Iron was a more useful material, especially for making hoes, spear shafts and the like. Moreover, outside highland areas (e.g. bUganda, Rwanda, Burundi), high temperatures and low humidity are available during part of the year, and using dry ground for kilns/forges to reach the temperatures needed for iron work was relatively common. Again, I refer you to that reference, Mining and Metallurgy in Negro Africa.

    Note that there were several centres of bronze and brass work. Most of these centres are associated with a few ethnic groups. Outside these areas, bronze and brass should be rare. It should not be hard to separate IQ by region of ancestry, although it may take a more refined study, as e.g. Lagos was a capital of Nigeria for a long time, and thus has people from other regions, Yoruba (like Lagos) and otherwise. Presumably, people not from such centres, whether Yoruba or not (in the case of Lagos), should not be subject to selective pressures for resistance to lead uptake. Another approach, on the assumption of substantial ancient bronze/brass work, is to check layers of the soil corresponding to various eras, for lead content, akin to what was deposited in Mongol China.
    , @Johan Meyer
    Perhaps I should go further. The fact that high temperatures are needed to smelt iron does not necessarily require a high IQ to observe and copy---the onus is on you to demonstrate as much. Despite apparently very low mean IQs, sub-Saharan Africans have maintained and developed the technology.

    Considering the difficulty of smelting iron in cold and wet environments, the choice of bronze and brass for useful items may have simply been a practical choice, requiring shorter smelting times and less fuel. Also, if the genes responsible for variation in IQ operate mainly by varying lead sensitivity (e.g. uptake), then one has a theory of how cold (and wet) winters may have given rise to the presence of such genes. The real intellectual challenge would be smelting iron in such cold and wet environments, thus leading to the association of intelligence and iron ages with cold and wet winter areas.

    I continue to emphasize wet, as the specific heat capacity of water, and the specific enthalpies of fusion (melting) and condensation (evaporation) are major barriers to reaching high temperatures.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  73. Jm8 says:
    @Jm8
    Iron working (and smelting) is not an easy thing (it's quite complex and difficult, and requiring high temperatures among other things—and further innovations like carbon steel were also invented among a few cultures e.g. the Haya tribe of West Tanzania) (and is evident early in S.E. and Nigeria about as Early as in Eurasia), and I think there is no reason to associate bronze more with intelligence.

    But anyway, bronze and brass working (as well as copper working), along with iron, were common all over West Africa and some parts of Central Africa (Including—in West Africa, where bronze and brass are common: the many Yoruba kingdoms: Ife, Owe, Ijebu, Owe, Oyo and other subgroups etc); the Igbo—such as at Igbo Ukwu—and other South East Nigerian groups—who produced several styles of bronze work over the centuries incl. that called the "Lower Niger Bronze Industry; the Edo of the Benin kingdom and related tribes; as well as those of Central and Northern Nigeria; the Ashanti/Akan—including for gold weights, sculptures, etc—, Ewe, Ga and other ethnic groups in Ghana—both North and South; the peoples of Dahomey; cultures of the Senegambia and Mali such as among many Mande speaking tribes—like the Soninke, Mandinka, Bambara etc‚—including those of Djenne/Jenne-Jeno and other city states along the Niger—there are several early-mid medieval bronzes surviving from there; the many tribes of the Voltaic regions in Burkina and Niger, and other areas. In Central Africa bronze, brass and copper were sometimes worked (though in Central Africa, copper was the more common) by cultures like the Bakongo (with their various subgroups), the Bakuba, the Baluba and Songye, and various other tribes in the Central Africa region (and some in the East Central African lakes region in the Uganda/Rwanda area. Some tribes in Eastern Africa (such as Tanzania also worked some of those metals. Bronze was also traditionally worked by cultures in Cameroon (like the kingdoms of the Bamileke, Bamoum, Bafut, etc, and tribes in Liberia/Sierra Leone. I know less about southern Africa (and bronze and brass may have been rarer in southern Africa).

    An overview—first link below (albeit somewhat brief, but inclusive) of the occurrence of bronze working from various bronze working cultures (with a focus on West Africa—where a great many tribes practiced it over large areas).

    http://www.sahistory.org.za/sites/default/files/DC/asjan59.17/asjan59.17.pdf


    A few examples:

    https://www.metmuseum.org/art/collection/search/310770
    https://www.metmuseum.org/art/collection/search/310257

    https://www.google.com/search?q=Ife+bronze&source=lnms&tbm=isch&sa=X&ved=0ahUKEwj4oL227u_aAhVxplkKHW9fBnoQ_AUICigB&biw=1015&bih=784

    Bamoum Bronze from Cameroon:

    gong at left link below w/ sculpted spider
    https://www.pinterest.com/pin/473933560769434257/

    https://auction.catawiki.com/kavels/9045487-large-bronze-pipe-bamun-bamileke-cameroon

    Edit: “…cultures of the Senegambia and Mali such as among many Mande speaking tribes—like the Soninke, Mandinka, Bambara etc‚—including those of Djenne/Jenne-Jeno and other city states along the Niger river—there are many early-mid medieval bronze sculptures, jewelry, and weapons/tools surviving from there”

    from Djenne:

    (1st link—should be image result at top toward right: abstract ornament with stylized pointing hands, outward hooks on either side, and small human figure in between)

    https://www.google.com/search?biw=1138&bih=885&tbm=isch&sa=1&ei=aVnuWuzoG-e1ggfn3ragDg&q=djenne+human+bronze+ornament&oq=djenne+human+bronze+ornament&gs_l=img.3…179415.180081.0.180277.6.6.0.0.0.0.77.363.6.6.0&#8230;.0…1c.1.64.img..0.0.0….0.WmVGhibkNHc#imgrc=cioRZjUKhG_TKM:

    https://www.metmuseum.org/art/collection/search/310001

    https://www.google.com/search?biw=1089&bih=794&tbm=isch&sa=1&ei=71buWtuYDoXk_Abiw6GABA&q=djenne+bronze+&oq=djenne+bronze+&gs_l=img.3…270436.270436.0.270555.1.1.0.0.0.0.66.66.1.1.0&#8230;.0…1c.1.64.img..0.0.0….0.84DAQiOp28Q#imgrc=rHijkUQ75jas5M:

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  74. @Jm8
    Iron working (and smelting) is not an easy thing (it's quite complex and difficult, and requiring high temperatures among other things—and further innovations like carbon steel were also invented among a few cultures e.g. the Haya tribe of West Tanzania) (and is evident early in S.E. and Nigeria about as Early as in Eurasia), and I think there is no reason to associate bronze more with intelligence.

    But anyway, bronze and brass working (as well as copper working), along with iron, were common all over West Africa and some parts of Central Africa (Including—in West Africa, where bronze and brass are common: the many Yoruba kingdoms: Ife, Owe, Ijebu, Owe, Oyo and other subgroups etc); the Igbo—such as at Igbo Ukwu—and other South East Nigerian groups—who produced several styles of bronze work over the centuries incl. that called the "Lower Niger Bronze Industry; the Edo of the Benin kingdom and related tribes; as well as those of Central and Northern Nigeria; the Ashanti/Akan—including for gold weights, sculptures, etc—, Ewe, Ga and other ethnic groups in Ghana—both North and South; the peoples of Dahomey; cultures of the Senegambia and Mali such as among many Mande speaking tribes—like the Soninke, Mandinka, Bambara etc‚—including those of Djenne/Jenne-Jeno and other city states along the Niger—there are several early-mid medieval bronzes surviving from there; the many tribes of the Voltaic regions in Burkina and Niger, and other areas. In Central Africa bronze, brass and copper were sometimes worked (though in Central Africa, copper was the more common) by cultures like the Bakongo (with their various subgroups), the Bakuba, the Baluba and Songye, and various other tribes in the Central Africa region (and some in the East Central African lakes region in the Uganda/Rwanda area. Some tribes in Eastern Africa (such as Tanzania also worked some of those metals. Bronze was also traditionally worked by cultures in Cameroon (like the kingdoms of the Bamileke, Bamoum, Bafut, etc, and tribes in Liberia/Sierra Leone. I know less about southern Africa (and bronze and brass may have been rarer in southern Africa).

    An overview—first link below (albeit somewhat brief, but inclusive) of the occurrence of bronze working from various bronze working cultures (with a focus on West Africa—where a great many tribes practiced it over large areas).

    http://www.sahistory.org.za/sites/default/files/DC/asjan59.17/asjan59.17.pdf


    A few examples:

    https://www.metmuseum.org/art/collection/search/310770
    https://www.metmuseum.org/art/collection/search/310257

    https://www.google.com/search?q=Ife+bronze&source=lnms&tbm=isch&sa=X&ved=0ahUKEwj4oL227u_aAhVxplkKHW9fBnoQ_AUICigB&biw=1015&bih=784

    Bamoum Bronze from Cameroon:

    gong at left link below w/ sculpted spider
    https://www.pinterest.com/pin/473933560769434257/

    https://auction.catawiki.com/kavels/9045487-large-bronze-pipe-bamun-bamileke-cameroon

    Some comments:

    If genes adding to intelligence enable bronze-work, then certainly the presence of the genes would have to precede the metal work. If, however, the genes are a response to an intelligence-destroying poison in the source material (soft metal deposits usually contain lead), then bronze work should precede the elevated presence of the genes. In this regard, genetically variable sensitivity to environment violates the assumption of separate (additive with separate terms) genetic and environmental contribution to intelligence, in which case it becomes meaningless to speak of x percent contribution of environment (or genes) to intelligence, as genes are implicated in the effect of environment in such models (dose response as a function of genes).

    Most of the objects are art, status objects or ritual objects. Tools or weapons are rare. You have not addressed the matter of wide-spread iron metallurgy, especially regarding the link that I posted. Iron was a more useful material, especially for making hoes, spear shafts and the like. Moreover, outside highland areas (e.g. bUganda, Rwanda, Burundi), high temperatures and low humidity are available during part of the year, and using dry ground for kilns/forges to reach the temperatures needed for iron work was relatively common. Again, I refer you to that reference, Mining and Metallurgy in Negro Africa.

    Note that there were several centres of bronze and brass work. Most of these centres are associated with a few ethnic groups. Outside these areas, bronze and brass should be rare. It should not be hard to separate IQ by region of ancestry, although it may take a more refined study, as e.g. Lagos was a capital of Nigeria for a long time, and thus has people from other regions, Yoruba (like Lagos) and otherwise. Presumably, people not from such centres, whether Yoruba or not (in the case of Lagos), should not be subject to selective pressures for resistance to lead uptake. Another approach, on the assumption of substantial ancient bronze/brass work, is to check layers of the soil corresponding to various eras, for lead content, akin to what was deposited in Mongol China.

    Read More
    • Replies: @Johan Meyer
    I should also note that some highland areas do have hot and relatively dry climates, e.g. much of Angola.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  75. @Johan Meyer
    Some comments:

    If genes adding to intelligence enable bronze-work, then certainly the presence of the genes would have to precede the metal work. If, however, the genes are a response to an intelligence-destroying poison in the source material (soft metal deposits usually contain lead), then bronze work should precede the elevated presence of the genes. In this regard, genetically variable sensitivity to environment violates the assumption of separate (additive with separate terms) genetic and environmental contribution to intelligence, in which case it becomes meaningless to speak of x percent contribution of environment (or genes) to intelligence, as genes are implicated in the effect of environment in such models (dose response as a function of genes).

    Most of the objects are art, status objects or ritual objects. Tools or weapons are rare. You have not addressed the matter of wide-spread iron metallurgy, especially regarding the link that I posted. Iron was a more useful material, especially for making hoes, spear shafts and the like. Moreover, outside highland areas (e.g. bUganda, Rwanda, Burundi), high temperatures and low humidity are available during part of the year, and using dry ground for kilns/forges to reach the temperatures needed for iron work was relatively common. Again, I refer you to that reference, Mining and Metallurgy in Negro Africa.

    Note that there were several centres of bronze and brass work. Most of these centres are associated with a few ethnic groups. Outside these areas, bronze and brass should be rare. It should not be hard to separate IQ by region of ancestry, although it may take a more refined study, as e.g. Lagos was a capital of Nigeria for a long time, and thus has people from other regions, Yoruba (like Lagos) and otherwise. Presumably, people not from such centres, whether Yoruba or not (in the case of Lagos), should not be subject to selective pressures for resistance to lead uptake. Another approach, on the assumption of substantial ancient bronze/brass work, is to check layers of the soil corresponding to various eras, for lead content, akin to what was deposited in Mongol China.

    I should also note that some highland areas do have hot and relatively dry climates, e.g. much of Angola.

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  76. @Jm8
    Iron working (and smelting) is not an easy thing (it's quite complex and difficult, and requiring high temperatures among other things—and further innovations like carbon steel were also invented among a few cultures e.g. the Haya tribe of West Tanzania) (and is evident early in S.E. and Nigeria about as Early as in Eurasia), and I think there is no reason to associate bronze more with intelligence.

    But anyway, bronze and brass working (as well as copper working), along with iron, were common all over West Africa and some parts of Central Africa (Including—in West Africa, where bronze and brass are common: the many Yoruba kingdoms: Ife, Owe, Ijebu, Owe, Oyo and other subgroups etc); the Igbo—such as at Igbo Ukwu—and other South East Nigerian groups—who produced several styles of bronze work over the centuries incl. that called the "Lower Niger Bronze Industry; the Edo of the Benin kingdom and related tribes; as well as those of Central and Northern Nigeria; the Ashanti/Akan—including for gold weights, sculptures, etc—, Ewe, Ga and other ethnic groups in Ghana—both North and South; the peoples of Dahomey; cultures of the Senegambia and Mali such as among many Mande speaking tribes—like the Soninke, Mandinka, Bambara etc‚—including those of Djenne/Jenne-Jeno and other city states along the Niger—there are several early-mid medieval bronzes surviving from there; the many tribes of the Voltaic regions in Burkina and Niger, and other areas. In Central Africa bronze, brass and copper were sometimes worked (though in Central Africa, copper was the more common) by cultures like the Bakongo (with their various subgroups), the Bakuba, the Baluba and Songye, and various other tribes in the Central Africa region (and some in the East Central African lakes region in the Uganda/Rwanda area. Some tribes in Eastern Africa (such as Tanzania also worked some of those metals. Bronze was also traditionally worked by cultures in Cameroon (like the kingdoms of the Bamileke, Bamoum, Bafut, etc, and tribes in Liberia/Sierra Leone. I know less about southern Africa (and bronze and brass may have been rarer in southern Africa).

    An overview—first link below (albeit somewhat brief, but inclusive) of the occurrence of bronze working from various bronze working cultures (with a focus on West Africa—where a great many tribes practiced it over large areas).

    http://www.sahistory.org.za/sites/default/files/DC/asjan59.17/asjan59.17.pdf


    A few examples:

    https://www.metmuseum.org/art/collection/search/310770
    https://www.metmuseum.org/art/collection/search/310257

    https://www.google.com/search?q=Ife+bronze&source=lnms&tbm=isch&sa=X&ved=0ahUKEwj4oL227u_aAhVxplkKHW9fBnoQ_AUICigB&biw=1015&bih=784

    Bamoum Bronze from Cameroon:

    gong at left link below w/ sculpted spider
    https://www.pinterest.com/pin/473933560769434257/

    https://auction.catawiki.com/kavels/9045487-large-bronze-pipe-bamun-bamileke-cameroon

    Perhaps I should go further. The fact that high temperatures are needed to smelt iron does not necessarily require a high IQ to observe and copy—the onus is on you to demonstrate as much. Despite apparently very low mean IQs, sub-Saharan Africans have maintained and developed the technology.

    Considering the difficulty of smelting iron in cold and wet environments, the choice of bronze and brass for useful items may have simply been a practical choice, requiring shorter smelting times and less fuel. Also, if the genes responsible for variation in IQ operate mainly by varying lead sensitivity (e.g. uptake), then one has a theory of how cold (and wet) winters may have given rise to the presence of such genes. The real intellectual challenge would be smelting iron in such cold and wet environments, thus leading to the association of intelligence and iron ages with cold and wet winter areas.

    I continue to emphasize wet, as the specific heat capacity of water, and the specific enthalpies of fusion (melting) and condensation (evaporation) are major barriers to reaching high temperatures.

    Read More
    • Replies: @Jm8
    I should mention that in that comment of mine to which you responded (in discussing possible "true iq numbers" of groups), I was comparing Africans to/discussing them in reference to Amerindians (who, in most cases, did not have metallurgy at all at any point in their history; in most cases—with some exceptions—neither iron nor bronze/brazz), rather than in reference to bronze-using Eurasians. So a discussion of the relative commonness of bronze in Africa vs Eurasia might not be so relevant.

    "Most of these centres are associated with a few ethnic groups."

    Not quite true. Quite a significant/substantial number of ethnic groups engaged in bronze work, though it may have been more common in some than others (relative "centers" as you say).

    "You have not addressed the matter of wide-spread iron metallurgy, especially regarding the link that I posted" ..."Iron was a more useful material, especially for making hoes, spear shafts and the like."

    Yes. Once iron metallurgy was invented, often replaced bronze (to a large degree) for certain functions. This occurred in most of Eurasia; when iron metallurgy was developed it was widely preferred in part because of its strength and hardness (and after the iron age bronze working was in general retained more for artistic objects). In much of Africa, iron was developed with little, sometimes no, preceding bronze phase—but bronze was present in many such regions not long after. Sometimes, after iron, the two metals coexisted, but with the presence of iron, bronze was used less for objects such as weapons/tools (though as in Eurasia, after its iron age, it was employed for, as you mention, art, status objects, ritual objects, jewelry—and at times practical objects such as weapons.

    I will look over your comments (and the points ideas therein—which do seem interesting, though I'm not sure the posited correlations you present are true, they could be true, and I will have to research the issue more) further (and will read the source you linked: "Mining and Metallurgy in Negro Africa") to see if I have more in the way of response.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  77. Jm8 says:
    @Johan Meyer
    Perhaps I should go further. The fact that high temperatures are needed to smelt iron does not necessarily require a high IQ to observe and copy---the onus is on you to demonstrate as much. Despite apparently very low mean IQs, sub-Saharan Africans have maintained and developed the technology.

    Considering the difficulty of smelting iron in cold and wet environments, the choice of bronze and brass for useful items may have simply been a practical choice, requiring shorter smelting times and less fuel. Also, if the genes responsible for variation in IQ operate mainly by varying lead sensitivity (e.g. uptake), then one has a theory of how cold (and wet) winters may have given rise to the presence of such genes. The real intellectual challenge would be smelting iron in such cold and wet environments, thus leading to the association of intelligence and iron ages with cold and wet winter areas.

    I continue to emphasize wet, as the specific heat capacity of water, and the specific enthalpies of fusion (melting) and condensation (evaporation) are major barriers to reaching high temperatures.

    I should mention that in that comment of mine to which you responded (in discussing possible “true iq numbers” of groups), I was comparing Africans to/discussing them in reference to Amerindians (who, in most cases, did not have metallurgy at all at any point in their history; in most cases—with some exceptions—neither iron nor bronze/brazz), rather than in reference to bronze-using Eurasians. So a discussion of the relative commonness of bronze in Africa vs Eurasia might not be so relevant.

    “Most of these centres are associated with a few ethnic groups.”

    Not quite true. Quite a significant/substantial number of ethnic groups engaged in bronze work, though it may have been more common in some than others (relative “centers” as you say).

    “You have not addressed the matter of wide-spread iron metallurgy, especially regarding the link that I posted” …”Iron was a more useful material, especially for making hoes, spear shafts and the like.”

    Yes. Once iron metallurgy was invented, often replaced bronze (to a large degree) for certain functions. This occurred in most of Eurasia; when iron metallurgy was developed it was widely preferred in part because of its strength and hardness (and after the iron age bronze working was in general retained more for artistic objects). In much of Africa, iron was developed with little, sometimes no, preceding bronze phase—but bronze was present in many such regions not long after. Sometimes, after iron, the two metals coexisted, but with the presence of iron, bronze was used less for objects such as weapons/tools (though as in Eurasia, after its iron age, it was employed for, as you mention, art, status objects, ritual objects, jewelry—and at times practical objects such as weapons.

    I will look over your comments (and the points ideas therein—which do seem interesting, though I’m not sure the posited correlations you present are true, they could be true, and I will have to research the issue more) further (and will read the source you linked: “Mining and Metallurgy in Negro Africa”) to see if I have more in the way of response.

    Read More
    • Replies: @Johan Meyer
    May I ask James Thompson to forward you a copy of my open letter to him and others? I was under the impression that a big motivation of this discussion was to see whether Piffer's equation/₥odel could be justified in terms of known (apparent) direct genetic contribution to IQ on an individual level, and in my open letter, I suggested an alternative possibility. At present, I lack funds to set up a website for hosting the open letter, and I would like to avoid too wide a distribution on account of current hysteria regarding HBD.
    , @Johan Meyer
    Briefly, my argument is as follows: bronze/brass work tends to produce lead poisoning, especially in antiquity. The per capita per annum amount of soft metal work determines overall lead poisoning in antiquity. Adult lead poisoning leads to reduced sperm quality, which then constitutes a selective pressure for reduced lead uptake per unit lead in the environment. Genes performing such functions (reducing lead uptake) are expected to produce high IQ populations in our era, due to leaded petrol/gasoline and paint, as certain ancient populations were subject to often far worse lead poisoning than what obtained in the last forty years. As environmental contribution of lead is highly variable over space and time, individuals with such genes, or without, may be subject to varying exposures to lead, thus reducing the correlation between individual IQ and such genes, while still allowing their effects to be seen in averages of large populations, as the standard error of mean lead poisoning in a population will be small.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  78. @Jm8
    I should mention that in that comment of mine to which you responded (in discussing possible "true iq numbers" of groups), I was comparing Africans to/discussing them in reference to Amerindians (who, in most cases, did not have metallurgy at all at any point in their history; in most cases—with some exceptions—neither iron nor bronze/brazz), rather than in reference to bronze-using Eurasians. So a discussion of the relative commonness of bronze in Africa vs Eurasia might not be so relevant.

    "Most of these centres are associated with a few ethnic groups."

    Not quite true. Quite a significant/substantial number of ethnic groups engaged in bronze work, though it may have been more common in some than others (relative "centers" as you say).

    "You have not addressed the matter of wide-spread iron metallurgy, especially regarding the link that I posted" ..."Iron was a more useful material, especially for making hoes, spear shafts and the like."

    Yes. Once iron metallurgy was invented, often replaced bronze (to a large degree) for certain functions. This occurred in most of Eurasia; when iron metallurgy was developed it was widely preferred in part because of its strength and hardness (and after the iron age bronze working was in general retained more for artistic objects). In much of Africa, iron was developed with little, sometimes no, preceding bronze phase—but bronze was present in many such regions not long after. Sometimes, after iron, the two metals coexisted, but with the presence of iron, bronze was used less for objects such as weapons/tools (though as in Eurasia, after its iron age, it was employed for, as you mention, art, status objects, ritual objects, jewelry—and at times practical objects such as weapons.

    I will look over your comments (and the points ideas therein—which do seem interesting, though I'm not sure the posited correlations you present are true, they could be true, and I will have to research the issue more) further (and will read the source you linked: "Mining and Metallurgy in Negro Africa") to see if I have more in the way of response.

    May I ask James Thompson to forward you a copy of my open letter to him and others? I was under the impression that a big motivation of this discussion was to see whether Piffer’s equation/₥odel could be justified in terms of known (apparent) direct genetic contribution to IQ on an individual level, and in my open letter, I suggested an alternative possibility. At present, I lack funds to set up a website for hosting the open letter, and I would like to avoid too wide a distribution on account of current hysteria regarding HBD.

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  79. @Jm8
    I should mention that in that comment of mine to which you responded (in discussing possible "true iq numbers" of groups), I was comparing Africans to/discussing them in reference to Amerindians (who, in most cases, did not have metallurgy at all at any point in their history; in most cases—with some exceptions—neither iron nor bronze/brazz), rather than in reference to bronze-using Eurasians. So a discussion of the relative commonness of bronze in Africa vs Eurasia might not be so relevant.

    "Most of these centres are associated with a few ethnic groups."

    Not quite true. Quite a significant/substantial number of ethnic groups engaged in bronze work, though it may have been more common in some than others (relative "centers" as you say).

    "You have not addressed the matter of wide-spread iron metallurgy, especially regarding the link that I posted" ..."Iron was a more useful material, especially for making hoes, spear shafts and the like."

    Yes. Once iron metallurgy was invented, often replaced bronze (to a large degree) for certain functions. This occurred in most of Eurasia; when iron metallurgy was developed it was widely preferred in part because of its strength and hardness (and after the iron age bronze working was in general retained more for artistic objects). In much of Africa, iron was developed with little, sometimes no, preceding bronze phase—but bronze was present in many such regions not long after. Sometimes, after iron, the two metals coexisted, but with the presence of iron, bronze was used less for objects such as weapons/tools (though as in Eurasia, after its iron age, it was employed for, as you mention, art, status objects, ritual objects, jewelry—and at times practical objects such as weapons.

    I will look over your comments (and the points ideas therein—which do seem interesting, though I'm not sure the posited correlations you present are true, they could be true, and I will have to research the issue more) further (and will read the source you linked: "Mining and Metallurgy in Negro Africa") to see if I have more in the way of response.

    Briefly, my argument is as follows: bronze/brass work tends to produce lead poisoning, especially in antiquity. The per capita per annum amount of soft metal work determines overall lead poisoning in antiquity. Adult lead poisoning leads to reduced sperm quality, which then constitutes a selective pressure for reduced lead uptake per unit lead in the environment. Genes performing such functions (reducing lead uptake) are expected to produce high IQ populations in our era, due to leaded petrol/gasoline and paint, as certain ancient populations were subject to often far worse lead poisoning than what obtained in the last forty years. As environmental contribution of lead is highly variable over space and time, individuals with such genes, or without, may be subject to varying exposures to lead, thus reducing the correlation between individual IQ and such genes, while still allowing their effects to be seen in averages of large populations, as the standard error of mean lead poisoning in a population will be small.

    Read More
    • Replies: @Jm8
    "May I ask James Thompson to forward you a copy of my open letter to him and others?"

    Yes, if you would like. I would be interested to read it.

    It may indeed be (it seems to me) that certain populations are more resistant to the iq-lowering effects of lead (which is what, it seems you are suggesting), due to a longer historical exposure to it (and it may be (likely is the case) that many Eurasian groups were exposed more for a longer time (with longer bronze ages—or with bronze ages, in/in contrast to the parts of Africa where bronze ages—in which bronze in the absence of iron was relied upon—did not occur) than some-many groups in areas of Africa, or the Americas (and elsewhere).

    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  80. Jm8 says:
    @Johan Meyer
    Briefly, my argument is as follows: bronze/brass work tends to produce lead poisoning, especially in antiquity. The per capita per annum amount of soft metal work determines overall lead poisoning in antiquity. Adult lead poisoning leads to reduced sperm quality, which then constitutes a selective pressure for reduced lead uptake per unit lead in the environment. Genes performing such functions (reducing lead uptake) are expected to produce high IQ populations in our era, due to leaded petrol/gasoline and paint, as certain ancient populations were subject to often far worse lead poisoning than what obtained in the last forty years. As environmental contribution of lead is highly variable over space and time, individuals with such genes, or without, may be subject to varying exposures to lead, thus reducing the correlation between individual IQ and such genes, while still allowing their effects to be seen in averages of large populations, as the standard error of mean lead poisoning in a population will be small.

    “May I ask James Thompson to forward you a copy of my open letter to him and others?”

    Yes, if you would like. I would be interested to read it.

    It may indeed be (it seems to me) that certain populations are more resistant to the iq-lowering effects of lead (which is what, it seems you are suggesting), due to a longer historical exposure to it (and it may be (likely is the case) that many Eurasian groups were exposed more for a longer time (with longer bronze ages—or with bronze ages, in/in contrast to the parts of Africa where bronze ages—in which bronze in the absence of iron was relied upon—did not occur) than some-many groups in areas of Africa, or the Americas (and elsewhere).

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

    Once again, in the spirit of the fearless examination of the intellect, I ask you to subject his work to merciless enquiry and savage criticism. Over to you.

    The best examples of critical arguments which I believe apply to Piffer’s work come from the Coop lab. Here are two examples (see this comment for more discussion).

    Pygmies and height PGS discussed in this preprint: https://www.biorxiv.org/content/early/2017/11/06/167551

    These polygenic scores should not be viewed as phenotypic predictions across populations. For example, the Maasai and Biaka pygmy populations have similar polygenic scores despite having dramatic differences in height.28 Discrepancies between polygenic scores and actual phenotypes may be expected to occur either because of purely environmental influences on phenotype, as well as gene-by-gene and gene-by-environment interactions. We also expect that the accuracy of these scores when viewed as predictions should decay with genetic distance from Europe (where the GWAS were carried out), due to changes in the structure of linkage disequilibrium (LD) between causal variants and tag SNPs picked up in GWAS, and because GWAS are biased toward discovering intermediate frequency variants, which will explain more variance in the region they are mapped in than outside of it. These caveats notwithstanding, the distribution of polygenic scores across populations can still be informative about the history of natural selection on a given phenotype,18 and a number of striking patterns are visible in their distribution. For example, there is a strong gradient in polygenic height scores running from east to west across Eurasia (Figure 1)

    And this blog post. Which I think is the best piece I have seen about this topic (PGS and their interpretation) including both rigorous science and a realistic look at possible issues with how results are interpreted. https://gcbias.org/2018/03/14/polygenic-scores-and-tea-drinking/

    In my opinion those critiques do not imply that Piffer’s work is wrong. More a caution about drawing premature conclusions (like the people who say there is no genetic component of between race differences in intelligence do ; ).

    What does everyone here think?

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter Display All Comments
  82. Factorize says:

    res, my economic development essay has finally been submitted and marked. Worked out great!
    The basic game plan was to talk about developing nations and intelligence. Here was the basic layout.
    I started with a fairly flaky– airy brief introduction about our wonderful approaching eugenic world.
    And then Bam-O– first sentence in the science section Nature Genetics: 3,000 SNPs for Educational Attainment– 37.6 SD of variants found. I then went through the basic findings about the polygenic architecture of intelligence and how this implies at least in theory massive potential IQs. I was told not to dwell on the science, so I presented enough of the outline to make it at least plausible. I then downshifted again and discussed high IQ populations and people and some of their development features. In the first part, I noted the magnification effect of shifting the Bell Curve to the right. When you shift the Bell Curve, even a fairly modest amount the number of extreme outliers greatly increase on the tail of the same side as the direction of the shift and greatly decrease on the other side. I then talked about von Neumann and what the world would be like if everyone had his level of cognitive ability. I finished up this section by talking of how everything appears to be lining up for a Genetic Singularity leading to the Singularity.

    The next section rotated into talking about how intelligence has large implications for significant social outcomes: The Life is an Intelligence Test (if you like it or not). Most of the research has been done in developed nations, so I had to present the evidence and leave it at that. I then noted the National IQ research that has found similar relationships in a broad range of nations, though using a less direct correlational approach. I did not want to become too bogged down with this material because this has already been extensively discussed and I wanted to move into some new ground.

    That was where my main section really shined. I talked about several aspects of human development such as income, demography, women, conflict … and how they relate to the psychometric point of view. But first to get the ball rolling I talked about Asian nations (specifically their industrial policies encouraging exports) as a nearly perfect demonstration of how critically important high g is in helping developing nations rapidly progress economically and socially. The literature talks about how entire nations become deeply interconnected networks of economic and social activity. Effectively coordinating such a complex system clearly requires a substantial amount of intelligence. Without such an effective guiding force, the entire economic engine can greatly underperform. From there I talked about the large income gains that would be expected with enhanced IQ, smart fractions and mentioned some nations that were notable for not living up to their GDP potential given their IQ. In the demography section, I noted how the world population graph can be thought of as measuring aggregate human psychometric potential. I know it is somewhat bogus, though I designated the summation of IQs across everyone at a given time as psi subscript H. I included a world population graph for the last 10,000 years and noted PSI H must almost certainly is now at or near the highest level ever achieved in human history. If you then also think more broadly and think of PSI of the universe (i.e., including computers) then we are now approaching the time in which computers will become the dominant force of psychometric potential of the universe. I then also talked about how merely scaling up Bell Curves leads to more people at the extreme tails and how this would be of relevance to developing nations. Specifically more people will due to the polygenic nature of IQ lead to a greater number of people with high IQs who can then create cognitive elites and possibly create a perpetuating high IQ fraction.

    It was all fairly standard. Yet, one notable spark I came up with was about dysgenics. Our textbook highlighted what was presented as a successful effort to reduce fertility in a developing nation. It described how an effort was launched in an area of high literacy and women empowerment to reduce fertility rates. This effort was noted as being nominally a highly successful example of family planning. Yet, it was also noted that in another region of this nation that did not have as high a rate of literacy or women empowerment family planning efforts had been much less successful (if they had even been tried at all). I tried my best to point out the glaring lack of psychometric understanding in this situation. Encouraging high literacy (i.e., high IQ) women to reduce their fertility while at the same time having much less success with low literacy (i.e., lower IQ) women to reduce their fertility could create a psychometric differential that could greatly destabilize a society.
    Of course, when you run these programs you will be motivated to show how successful you have been at hitting some top line reduction in fertility, and so it seems likely that one would try to invest one’s efforts with those most likely to respond to the message (i.e., the most intelligent). Apparently
    modern theory is able to effectively model the fertility decision making process and a range of interventions are available to greatly reduce fertility (through education, jobs, net price of children etc.). I noted that for some developing nations shifting down their Bell Curve could result in moving below some psychometric threshold and possibly a social catastrophe ensuing ( from the magnification effect noted above). This was meant as a cautionary tale in that simply ignoring psychometrics does not magically mean the consequences will disappear. Clearly some highly idealistic people could do a great deal of harm to very vulnerable societies if they did not think carefully about what they were doing.

    Read More
    • Replies: @res

    Clearly some highly idealistic people could do a great deal of harm to very vulnerable societies if they did not think carefully about what they were doing.
     
    Indeed. Help the least able to reproduce more and discourage the most able from reproducing. Sounds rather like the US.

    How was your essay received? Did it spark any discussion?
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  83. res says:
    @Factorize
    res, my economic development essay has finally been submitted and marked. Worked out great!
    The basic game plan was to talk about developing nations and intelligence. Here was the basic layout.
    I started with a fairly flaky-- airy brief introduction about our wonderful approaching eugenic world.
    And then Bam-O-- first sentence in the science section Nature Genetics: 3,000 SNPs for Educational Attainment-- 37.6 SD of variants found. I then went through the basic findings about the polygenic architecture of intelligence and how this implies at least in theory massive potential IQs. I was told not to dwell on the science, so I presented enough of the outline to make it at least plausible. I then downshifted again and discussed high IQ populations and people and some of their development features. In the first part, I noted the magnification effect of shifting the Bell Curve to the right. When you shift the Bell Curve, even a fairly modest amount the number of extreme outliers greatly increase on the tail of the same side as the direction of the shift and greatly decrease on the other side. I then talked about von Neumann and what the world would be like if everyone had his level of cognitive ability. I finished up this section by talking of how everything appears to be lining up for a Genetic Singularity leading to the Singularity.

    The next section rotated into talking about how intelligence has large implications for significant social outcomes: The Life is an Intelligence Test (if you like it or not). Most of the research has been done in developed nations, so I had to present the evidence and leave it at that. I then noted the National IQ research that has found similar relationships in a broad range of nations, though using a less direct correlational approach. I did not want to become too bogged down with this material because this has already been extensively discussed and I wanted to move into some new ground.

    That was where my main section really shined. I talked about several aspects of human development such as income, demography, women, conflict ... and how they relate to the psychometric point of view. But first to get the ball rolling I talked about Asian nations (specifically their industrial policies encouraging exports) as a nearly perfect demonstration of how critically important high g is in helping developing nations rapidly progress economically and socially. The literature talks about how entire nations become deeply interconnected networks of economic and social activity. Effectively coordinating such a complex system clearly requires a substantial amount of intelligence. Without such an effective guiding force, the entire economic engine can greatly underperform. From there I talked about the large income gains that would be expected with enhanced IQ, smart fractions and mentioned some nations that were notable for not living up to their GDP potential given their IQ. In the demography section, I noted how the world population graph can be thought of as measuring aggregate human psychometric potential. I know it is somewhat bogus, though I designated the summation of IQs across everyone at a given time as psi subscript H. I included a world population graph for the last 10,000 years and noted PSI H must almost certainly is now at or near the highest level ever achieved in human history. If you then also think more broadly and think of PSI of the universe (i.e., including computers) then we are now approaching the time in which computers will become the dominant force of psychometric potential of the universe. I then also talked about how merely scaling up Bell Curves leads to more people at the extreme tails and how this would be of relevance to developing nations. Specifically more people will due to the polygenic nature of IQ lead to a greater number of people with high IQs who can then create cognitive elites and possibly create a perpetuating high IQ fraction.

    It was all fairly standard. Yet, one notable spark I came up with was about dysgenics. Our textbook highlighted what was presented as a successful effort to reduce fertility in a developing nation. It described how an effort was launched in an area of high literacy and women empowerment to reduce fertility rates. This effort was noted as being nominally a highly successful example of family planning. Yet, it was also noted that in another region of this nation that did not have as high a rate of literacy or women empowerment family planning efforts had been much less successful (if they had even been tried at all). I tried my best to point out the glaring lack of psychometric understanding in this situation. Encouraging high literacy (i.e., high IQ) women to reduce their fertility while at the same time having much less success with low literacy (i.e., lower IQ) women to reduce their fertility could create a psychometric differential that could greatly destabilize a society.
    Of course, when you run these programs you will be motivated to show how successful you have been at hitting some top line reduction in fertility, and so it seems likely that one would try to invest one's efforts with those most likely to respond to the message (i.e., the most intelligent). Apparently
    modern theory is able to effectively model the fertility decision making process and a range of interventions are available to greatly reduce fertility (through education, jobs, net price of children etc.). I noted that for some developing nations shifting down their Bell Curve could result in moving below some psychometric threshold and possibly a social catastrophe ensuing ( from the magnification effect noted above). This was meant as a cautionary tale in that simply ignoring psychometrics does not magically mean the consequences will disappear. Clearly some highly idealistic people could do a great deal of harm to very vulnerable societies if they did not think carefully about what they were doing.

    Clearly some highly idealistic people could do a great deal of harm to very vulnerable societies if they did not think carefully about what they were doing.

    Indeed. Help the least able to reproduce more and discourage the most able from reproducing. Sounds rather like the US.

    How was your essay received? Did it spark any discussion?

    Read More
    • Replies: @Factorize
    res, I would live to hear your thoughts about the new age of learning. I have noticed a dramatic increase in my ability to produce high quality academic work as the technology tools have continued to increase. What I find quite surprising is how very quiet the psychometric community seems to have been on this point.
    Educators must have very good insight into this question.

    Imagine the assignment that a primary school aged student likely submitted 20 years ago. They might have had access to a school library of a few thousand books. It is even possible that they would submit their assignment written with crayons. Consider what that assignment that same child might submit today. They might be granted access to tens of thousands of research jounals through their school and would likely be able to submit an electronic document that was of very high production quality.

    It is truly amazing! We have entered an era of extraordinary potential! Even those not much out of kindergarten have the tools that they need to make substantial contributions to the dialogue. I am sure if I were back there now that is exactly what I would be doing. I am not sure why more has not been said about this. We have truly enjoyed a new era in which very high achievement levels could be present almost anywhere.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  84. Factorize says:

    res, yes it really is not an overly inspiring concept for the big antlers in the psychometric community, though it felt enormously brilliant within economic development. Economic development is aggressively anti-psychometric. Our textbook went out of its way to mention a psychometric interpretation only to question something that is factually proven (namely,smarter mothers have smarter children).

    What I found so startling about the fertility planning writeup in the text was the complete lack of awareness of what the psychometric implications of dysgenic fertility would be in a developing nation. The chapter described how a new and more powerful understanding of what drives fertility has emerged over the last few decades. It is now highly possible to control fertility quietly with a fair amount of precision in many developing nations. Notable examples of this new found technodemographic control can be seen in nations such as Afghanistan and Yemen which have experienced possibly the most rapid fertility collapses ever recorded. Glaringly one of the variables noted in the course materials controlling fertility was the education of mothers. At no time was a direct link of education and intelligence ever made.

    When I read the section that described a nominally successful family planning program in a developing nation that rapidly pushed the high IQ region below replacement fertility while the
    described “backward” region maintained an extremely high fertility, I had one of those chewing gum popping moments. Um, did I read that one right? They weaponized psychometric collapse in
    a developing nation… and that’s a good thing? I suppose they hold galas to celebrate these successes of Western NGOs that are making a difference in the world. Ironically, of course, they are making a difference by rowing in the wrong direction.

    For some developing nations with IQs hovering around 70, such dysgenic fertility could push them
    beyond their psychometric breaking point. There is an enormous magnification effect that occurs in the direction of the intelligence change. Such psychometric collapse is all the more plausible when one realizes that in many of these nations upwards of half or more of their cognitive elite take the first plane out once they graduate. It is even more troubling because the psychometric research infrastructure does not exist in many of these nations. I believe you saw the recent article where they were able to prove using genetic samples from Iceland that this population is clearly in psychometric decline. The authors noted that a crisis might arise there over the next few centuries if this decline were to continue. At least, we know. However, for a continent of Africa I am not sure if any large scale psychometric GWAS have been done. Nations there could be right at a critical intelligence threshold and there might be no obvious indications that danger was present.

    Essay was very warmly received. I had been building momentum through the course as I learned what I needed to do to increase my grades. Everything was perfectly setup, so that my final assignment would be highly regarded. It was meant to be a 2,000 word essay, though I blew past the word limit by 8,000 words and I still had to leave out about half of my material.

    Great thinking about the discussion idea. I have thought that in the modern era of education that this would make such a great deal of sense. The entire notion that only markers and highly selected others would ever read academic output seems prehistoric. Opening up wider discussion of academic output in high schools etc. could add a tremendous vitality and focus to what can often seem pedantic and unrelated to life course work. I suppose that there would be a fair few young social conscience warriors that might see the question of protesting organizations that blindly encouraged fertility planning that could lead to the psychometric collapse of vulnerable developing nations as a moral duty. However, topics such as psychometrics have essentially become forbidden topics in academia. Perhaps we are already past the point where meaningful dialogue can even occur. This obviously and probably inevitably leads us to the dangerous circumstance in which dysgenic family planning in developing nations is celebrated.

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter Display All Comments
  85. res says:
    @res
    I am not an expert here so would like to hear other opinions, but I think the Racimo approach is very different from Piffer's.

    Piffer's approach is much more direct and has the virtue of being testable with the correlations he observes.

    Racimo's approach is much more complex and has the virtue of providing details of trait selection throughout population branching. I find it fascinating that we are able to infer this.

    It is worth noting that if Piffer's hypothesis about the initially detected SNPs being representative of overall selection pressure holds then I think Racimo's approach should be informative with a fairly small number of SNPs. Similarly, I think Racimo's work is supportive of Piffer's hypothesis and could be even more so if they did comparisons between different sizes of SNP sets.

    Figure 7 on page 19 is an interesting look at allele frequencies and effect size. Here is an earlier version:
    https://2.bp.blogspot.com/-VbM059x_Zf4/WTf9mVG4lHI/AAAAAAAASfE/W8jvgYQy0c4xkmhU-2gFQMByhobyIypZgCLcB/s400/Screen%2BShot%2B2017-06-07%2Bat%2B9.19.47%2BAM.png

    I think Figure S38 on page 64/87 is the money figure for the paper (not sure why they used Figure 8 in the body instead). S36 and S37 are similar. It shows selection on the branches of the world population tree for the four traits. The results seem intuitive to me except I am a bit surprised there is no sign of positive selection on the r-q population branch for EA. It is possible that the SNPs underlying that selection (if it exists) would need an EA GWAS including both Europeans and Africans to detect.

    This link is different, but gives an idea of what the four S38 panels look like (I think the S38 height version is better because it also shows negative selection for height in Asia):

    https://github.com/FerRacimo/PolyGraph/raw/master/HEIGHT_1KG_YRI_CEU_CHB_PEL_CLM.png

    Racimo et al. provide R code for height at https://github.com/FerRacimo/PolyGraph
    It would be interesting to try the following experiments:
    - Replace their height SNPs with the compressed sensing version.
    - Replace their height SNPs with recent EA/IQ results.

    I need to try playing with that. Though perhaps using it with many SNP results is a bad idea: "The trace from the MCMC run (slower to compute; may take hours or days, depending on the complexity of the graph and the number of trait-affecting SNPs)" They used 532 height SNPs: https://github.com/FerRacimo/PolyGraph/blob/master/GWAS_HEIGHT_1000genomes_allpops.txt

    Thinking about it some more, I think the experiment to try is running their analysis on Piffer's initial (small) IQ/EA SNP sets.

    P.S. Dr. Thompson, not sure what you meant in comment 35, but the second link Steve gave has full text of the preprint (all 107 pages!).

    I was looking through Racimo’s paper again today and took a closer look at two very interesting supplemental graphics.

    Figure S56 on page 105 is a variant of the idea for validating Piffer’s method I have discussed where one looks at successively more comprehensive sets of SNPs with a PGS. The figure gives genetic scores by 1000 Genomes population for the UKBB for p-values of 10^-9, 10^-8, and 10^-7. Visually we see good correspondence with some variation. It should be possible to match those with the phenotypic country values in a fashion similar to Piffer’s and check how the correlations vary.

    Perhaps a better variation would be to look at mutually exclusive groups of SNPs (e.g. bucket by p-value) in a similar fashion. This would prevent the issue of the intercorrelations of variables like A, A+B, A+B+C.

    Figure S58 compares graphics built using two different selection measures. IIRC their git code uses the qb statistics to initialize their MCMC calculations of the alpha parameters, but I am not clear on the relative strengths and weaknesses of the estimates.
    The qb statistics show higher levels of selection signals than the alpha parameters (and in a way that matches intuition for S58 height). I am guessing they are being conservative by showing the alpha parameters (in most/all? of the other graphics), but the qb statistics seem more informative and are much easier/faster to calculate.

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

    Clearly some highly idealistic people could do a great deal of harm to very vulnerable societies if they did not think carefully about what they were doing.
     
    Indeed. Help the least able to reproduce more and discourage the most able from reproducing. Sounds rather like the US.

    How was your essay received? Did it spark any discussion?

    res, I would live to hear your thoughts about the new age of learning. I have noticed a dramatic increase in my ability to produce high quality academic work as the technology tools have continued to increase. What I find quite surprising is how very quiet the psychometric community seems to have been on this point.
    Educators must have very good insight into this question.

    Imagine the assignment that a primary school aged student likely submitted 20 years ago. They might have had access to a school library of a few thousand books. It is even possible that they would submit their assignment written with crayons. Consider what that assignment that same child might submit today. They might be granted access to tens of thousands of research jounals through their school and would likely be able to submit an electronic document that was of very high production quality.

    It is truly amazing! We have entered an era of extraordinary potential! Even those not much out of kindergarten have the tools that they need to make substantial contributions to the dialogue. I am sure if I were back there now that is exactly what I would be doing. I am not sure why more has not been said about this. We have truly enjoyed a new era in which very high achievement levels could be present almost anywhere.

    Read More
    • Replies: @James Thompson
    Well, I had tools available but did not use them, because they were very hard for me to master. I doubt there will be the big increase you anticipate, but it would be good to be wrong.
    , @res

    I have noticed a dramatic increase in my ability to produce high quality academic work as the technology tools have continued to increase.
     
    I see the same thing. For me it is the combination of access to research materials (e.g. books and papers) and analytical data/tools. I think this is a combination of increased accessibility along with an increased (and increasing) volume of new information. Worth mentioning the increase in computer power driving analytical abilities as well.

    submit an electronic document that was of very high production quality.
     
    There is a very real risk of form overshadowing function here. A thoughtful and insightful assignment written with crayon is preferable to a pretty, well referenced, and even well written piece with negative intellectual content (cf. IQ debunking articles in places like the NYT).

    I am not sure why more has not been said about this.
     
    I get the sense you are in college now (feel free to correct me). I think part of this is just where we are in the hype cycle for this phenomenon: https://en.wikipedia.org/wiki/Hype_cycle
    I think we had the "world's knowledge on your desktop" being overhyped through the early internet age and now it just goes unsaid. There might be some disillusionment with realizing that despite the availability of information our abilities to use it well (as a society, especially in the public sphere) don't always keep up.

    We have truly enjoyed a new era in which very high achievement levels could be present almost anywhere.
     
    It would be interesting to see the temporal trend characterized in a Charles Murray's Human Accomplishment sense. It is hard to distinguish quality from quantity without the benefit of hindsight though.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  87. @Factorize
    res, I would live to hear your thoughts about the new age of learning. I have noticed a dramatic increase in my ability to produce high quality academic work as the technology tools have continued to increase. What I find quite surprising is how very quiet the psychometric community seems to have been on this point.
    Educators must have very good insight into this question.

    Imagine the assignment that a primary school aged student likely submitted 20 years ago. They might have had access to a school library of a few thousand books. It is even possible that they would submit their assignment written with crayons. Consider what that assignment that same child might submit today. They might be granted access to tens of thousands of research jounals through their school and would likely be able to submit an electronic document that was of very high production quality.

    It is truly amazing! We have entered an era of extraordinary potential! Even those not much out of kindergarten have the tools that they need to make substantial contributions to the dialogue. I am sure if I were back there now that is exactly what I would be doing. I am not sure why more has not been said about this. We have truly enjoyed a new era in which very high achievement levels could be present almost anywhere.

    Well, I had tools available but did not use them, because they were very hard for me to master. I doubt there will be the big increase you anticipate, but it would be good to be wrong.

    Read More
    • Replies: @res
    I imagine you have had a good look at the tool capabilities over the years. Could you talk a bit about how things have changed? My sense is data is much more available and the tools are much more powerful, but I don't know how much usability has changed.

    I think you draw out a key issue with Factorize's optimism. Namely the desire and ability to:
    1. Master the relevant areas of knowledge enough to first understand and later add original extensions.
    2. Master the tools required to accomplish this.
    Are both limited. In addition, at a country (or so, not sure which level exactly) level there may be issues with physical infrastructure, institutions, and overall culture helping or hindering knowledge discovery.
    , @Factorize
    I was speaking somewhat from personal experience. When I had been asked to do assignments before the age of infotech, I was nearly completely stumped. I had absolutely no idea how to complete such assignments; several times I gave up in frustration without even making a submission. However, with this new era of information technology, the difference for me is dramatic. I do not want to brag (merely state the facts), though in my last assignment I scored a 100% and it included PAGES of references. In assignments that I submitted before the information age, the references section would be more accurately titled Reference. I have the impression that the assignments are now not always even been read and a very high mark is being given as a matter of course.

    My latest assignment has set a new all time high in the quality of my submitted work. In fact one of the things that I noticed was that as this was an economics course, they had not seemed overly interested in strictly enforcing APA style referencing. Some of the non-conventional referencing in my previously submitted assignments for the course had went without comment. This allowed me to focus more on higher level aspects of the writing process and I was able to think up a fairly interesting rhetorical hook involving von Neumann and the opening of A Tale of Two Cities. What was especially interesting was that in a previous psychology course there was this endless nitpicking about commas etc. in the APA reference section so I was never able to focus on the more important writing elements to produce a great assignment. I have contacted my online university about whether they could provide even more tools such as APA citation, plagiarism and virtual writing coaches that would allow students to become better writers and not better copy editors.

    I know that if I could time transport myself forward from kindergarten to the present that I would use the existing technology to produce equally extreme academic output throughout my academic career. While this might not be true for everyone, I know that it would be true for me.

    It does, however, greatly surprise me that this observation would even be challenged. I think what happens so many times is that we pass through some very large technological change and all the boats rise in the water and for whatever reason nobody even bothers noticing that the entire landscape has shifted. As noted above, for me this is not a little, but a very very large shift. Admittedly this is not actually a g shift, though in terms of my measured academic performance
    the change is quite dramatic. I have kept previous assignments and it is all too obvious which assignments would have been submitted in which era simply the quality of the work.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  88. res says:
    @Factorize
    res, I would live to hear your thoughts about the new age of learning. I have noticed a dramatic increase in my ability to produce high quality academic work as the technology tools have continued to increase. What I find quite surprising is how very quiet the psychometric community seems to have been on this point.
    Educators must have very good insight into this question.

    Imagine the assignment that a primary school aged student likely submitted 20 years ago. They might have had access to a school library of a few thousand books. It is even possible that they would submit their assignment written with crayons. Consider what that assignment that same child might submit today. They might be granted access to tens of thousands of research jounals through their school and would likely be able to submit an electronic document that was of very high production quality.

    It is truly amazing! We have entered an era of extraordinary potential! Even those not much out of kindergarten have the tools that they need to make substantial contributions to the dialogue. I am sure if I were back there now that is exactly what I would be doing. I am not sure why more has not been said about this. We have truly enjoyed a new era in which very high achievement levels could be present almost anywhere.

    I have noticed a dramatic increase in my ability to produce high quality academic work as the technology tools have continued to increase.

    I see the same thing. For me it is the combination of access to research materials (e.g. books and papers) and analytical data/tools. I think this is a combination of increased accessibility along with an increased (and increasing) volume of new information. Worth mentioning the increase in computer power driving analytical abilities as well.

    submit an electronic document that was of very high production quality.

    There is a very real risk of form overshadowing function here. A thoughtful and insightful assignment written with crayon is preferable to a pretty, well referenced, and even well written piece with negative intellectual content (cf. IQ debunking articles in places like the NYT).

    I am not sure why more has not been said about this.

    I get the sense you are in college now (feel free to correct me). I think part of this is just where we are in the hype cycle for this phenomenon: https://en.wikipedia.org/wiki/Hype_cycle
    I think we had the “world’s knowledge on your desktop” being overhyped through the early internet age and now it just goes unsaid. There might be some disillusionment with realizing that despite the availability of information our abilities to use it well (as a society, especially in the public sphere) don’t always keep up.

    We have truly enjoyed a new era in which very high achievement levels could be present almost anywhere.

    It would be interesting to see the temporal trend characterized in a Charles Murray’s Human Accomplishment sense. It is hard to distinguish quality from quantity without the benefit of hindsight though.

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  89. res says:
    @James Thompson
    Well, I had tools available but did not use them, because they were very hard for me to master. I doubt there will be the big increase you anticipate, but it would be good to be wrong.

    I imagine you have had a good look at the tool capabilities over the years. Could you talk a bit about how things have changed? My sense is data is much more available and the tools are much more powerful, but I don’t know how much usability has changed.

    I think you draw out a key issue with Factorize’s optimism. Namely the desire and ability to:
    1. Master the relevant areas of knowledge enough to first understand and later add original extensions.
    2. Master the tools required to accomplish this.
    Are both limited. In addition, at a country (or so, not sure which level exactly) level there may be issues with physical infrastructure, institutions, and overall culture helping or hindering knowledge discovery.

    Read More
    • Replies: @Factorize
    res, thank you for picking up on this idea. I think that once again when everyone does better there is often a surprising lack of interest in investigating what might have happened.

    One genetic feature that likely is driving my response is that I tend to believe that I have obsessional-compulsive type trait. My behavior of being all or nothing that I demonstrated with my assignments would tend to confirm my hunch. I expect that in this golden age of information having such a tendency would probably result in very extreme achievement in those with such a trait.

    It should be noted that the informational resources that I can access through my personal computer from my online university are considerable. When I enroll in a course there are over 30,000 reference journals that can be found in the virtual library. It is quite impressive. This could be provided to students anywhere in the world who had internet access without respect to the per capita GDP of their nation etc. It would be a fantastic way to leapfrog past the enormous expense that has been required to build up the cognitive resource base of developed nations. I am not sure how many physical university libraries would have such an extensive collection. I can only hope that similar access is now offered to all students at any level in their education. It makes a tremendous difference when you can access the same material that might be written from even a slightly different perspective. I have noticed that if you only read one account of a subject that you will often miss out important aspects of the subject material.

    What I find especially amusing about this is that this is one of the few questions in psychometrics that could easily be demonstrated in an experimental setting any time into the future, no matter how elaborate infotech becomes. All you would need to do is put a test subject in a environment for say a day and ask them to answer some research question: in one setting (control) they would only have access to say a high school library, in another setting (treatment) they would have access to a computer would enhanced internet access (i.e., research journals , ebooks). It should be obvious that with internet access the measured performance of the work would be enormously higher than with only a library.

    It is surprising to me that more has not been said about this. It is as if we superturbocharged the cognitive space and no one thought this was worth commenting on. While those on this blog are highly psychometrically aware and will likely call me on this, but I have to say this new internet environment has probably added about 30 IQ points to the quality of assignment that I can now submit. Up to this point there has not been a single question on any of my assignments that I could not research endlessly. In one of my assignments I was able to simply google the question and a quote on this exact question popped up from a Nobel winning scientist. Without the internet there would have been no possible way that I could have found this quote out of the tens of millions of books in print. The internet has quite clearly logarithmically amplified my intellectual reach.

    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  90. Factorize says:
    @James Thompson
    Well, I had tools available but did not use them, because they were very hard for me to master. I doubt there will be the big increase you anticipate, but it would be good to be wrong.

    I was speaking somewhat from personal experience. When I had been asked to do assignments before the age of infotech, I was nearly completely stumped. I had absolutely no idea how to complete such assignments; several times I gave up in frustration without even making a submission. However, with this new era of information technology, the difference for me is dramatic. I do not want to brag (merely state the facts), though in my last assignment I scored a 100% and it included PAGES of references. In assignments that I submitted before the information age, the references section would be more accurately titled Reference. I have the impression that the assignments are now not always even been read and a very high mark is being given as a matter of course.

    My latest assignment has set a new all time high in the quality of my submitted work. In fact one of the things that I noticed was that as this was an economics course, they had not seemed overly interested in strictly enforcing APA style referencing. Some of the non-conventional referencing in my previously submitted assignments for the course had went without comment. This allowed me to focus more on higher level aspects of the writing process and I was able to think up a fairly interesting rhetorical hook involving von Neumann and the opening of A Tale of Two Cities. What was especially interesting was that in a previous psychology course there was this endless nitpicking about commas etc. in the APA reference section so I was never able to focus on the more important writing elements to produce a great assignment. I have contacted my online university about whether they could provide even more tools such as APA citation, plagiarism and virtual writing coaches that would allow students to become better writers and not better copy editors.

    I know that if I could time transport myself forward from kindergarten to the present that I would use the existing technology to produce equally extreme academic output throughout my academic career. While this might not be true for everyone, I know that it would be true for me.

    It does, however, greatly surprise me that this observation would even be challenged. I think what happens so many times is that we pass through some very large technological change and all the boats rise in the water and for whatever reason nobody even bothers noticing that the entire landscape has shifted. As noted above, for me this is not a little, but a very very large shift. Admittedly this is not actually a g shift, though in terms of my measured academic performance
    the change is quite dramatic. I have kept previous assignments and it is all too obvious which assignments would have been submitted in which era simply the quality of the work.

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  91. Factorize says:
    @res
    I imagine you have had a good look at the tool capabilities over the years. Could you talk a bit about how things have changed? My sense is data is much more available and the tools are much more powerful, but I don't know how much usability has changed.

    I think you draw out a key issue with Factorize's optimism. Namely the desire and ability to:
    1. Master the relevant areas of knowledge enough to first understand and later add original extensions.
    2. Master the tools required to accomplish this.
    Are both limited. In addition, at a country (or so, not sure which level exactly) level there may be issues with physical infrastructure, institutions, and overall culture helping or hindering knowledge discovery.

    res, thank you for picking up on this idea. I think that once again when everyone does better there is often a surprising lack of interest in investigating what might have happened.

    One genetic feature that likely is driving my response is that I tend to believe that I have obsessional-compulsive type trait. My behavior of being all or nothing that I demonstrated with my assignments would tend to confirm my hunch. I expect that in this golden age of information having such a tendency would probably result in very extreme achievement in those with such a trait.

    It should be noted that the informational resources that I can access through my personal computer from my online university are considerable. When I enroll in a course there are over 30,000 reference journals that can be found in the virtual library. It is quite impressive. This could be provided to students anywhere in the world who had internet access without respect to the per capita GDP of their nation etc. It would be a fantastic way to leapfrog past the enormous expense that has been required to build up the cognitive resource base of developed nations. I am not sure how many physical university libraries would have such an extensive collection. I can only hope that similar access is now offered to all students at any level in their education. It makes a tremendous difference when you can access the same material that might be written from even a slightly different perspective. I have noticed that if you only read one account of a subject that you will often miss out important aspects of the subject material.

    What I find especially amusing about this is that this is one of the few questions in psychometrics that could easily be demonstrated in an experimental setting any time into the future, no matter how elaborate infotech becomes. All you would need to do is put a test subject in a environment for say a day and ask them to answer some research question: in one setting (control) they would only have access to say a high school library, in another setting (treatment) they would have access to a computer would enhanced internet access (i.e., research journals , ebooks). It should be obvious that with internet access the measured performance of the work would be enormously higher than with only a library.

    It is surprising to me that more has not been said about this. It is as if we superturbocharged the cognitive space and no one thought this was worth commenting on. While those on this blog are highly psychometrically aware and will likely call me on this, but I have to say this new internet environment has probably added about 30 IQ points to the quality of assignment that I can now submit. Up to this point there has not been a single question on any of my assignments that I could not research endlessly. In one of my assignments I was able to simply google the question and a quote on this exact question popped up from a Nobel winning scientist. Without the internet there would have been no possible way that I could have found this quote out of the tens of millions of books in print. The internet has quite clearly logarithmically amplified my intellectual reach.

    Read More
    • Replies: @res

    new internet environment has probably added about 30 IQ points to the quality of assignment that I can now submit.
     
    This is an interesting point (though I am not sure the unit of measurement is appropriate). The areas where I see the biggest improvements are:
    - Professional presentation. Formatting, adherence to guidelines (like your APA style comments above), spell checking.
    - Reference availability and searchability as you have discussed.
    - Data and analytics. And graphical presentation of data.
    - Source control. Being able to keep good records of changes to data, analytics, and presentation is incredibly helpful.
    - Replicability. I can hand someone an R markdown file and my data and they can reproduce (and extend) my results exactly. In a well formatted fashion.

    I find the combination of R, RStudio, R Markdown, and https://rpubs.com/ incredibly powerful as an environment for accomplishing all of the above. Davide Piffer's work provides a good example of what is possible. Both with RPubs: http://rpubs.com/Daxide/ and OSF: https://osf.io/ewxqj/
    Emil's website also has good examples: http://emilkirkegaard.dk/en/

    But none of those are more important than the ability to ask good questions, think clearly about them, draw useful conclusions, and express both the reasoning and the conclusions clearly and in a well supported fashion to others. The new tools just make doing so much easier--if you have the ability. Though one can argue the tools also make it easier to acquire the ability (the barriers to entry are lower).

    I wonder how all of this appears to professors critiquing student assignments. Are the improvements real, or are they just "turd polishing"? If you have any older professors (or if any here want to comment) perhaps you could ask them about this?

    there has not been a single question on any of my assignments that I could not research endlessly.
     
    The type of person who has a compulsion to do this as things interest them (me too) is exactly the type of person who thrives on this availability. I think many just find it overwhelming (as I do sometimes). I am not sure the improvement has been as large for the dedicated academic researcher (I would be interested in hearing counterpoints). For decades there have been libraries, books, journals, and photocopiers. And I think academics are (and have always been) quite good at accumulating the resources they need as individuals, labs, and fields.

    Without the internet there would have been no possible way that I could have found this quote out of the tens of millions of books in print.
     
    It is daunting to think how much harder things like that used to be. Even with access to a high end reference library. The ability we have now to find information and then track it back through a chain of references is mind boggling. Then there is the ability to link directly to those references either in whole or part (I find selected figures particular informative, not sure how others feel about that ; ).

    It is important to note that it is not just "the internet." It is the contents as well as the connectivity. Years ago it seemed like "everything is available" until you actually went and looked. Now almost everything really is available electronically even if not all of it is public. Getting all those books and papers digitized has taken time.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  92. res says:
    @Factorize
    res, thank you for picking up on this idea. I think that once again when everyone does better there is often a surprising lack of interest in investigating what might have happened.

    One genetic feature that likely is driving my response is that I tend to believe that I have obsessional-compulsive type trait. My behavior of being all or nothing that I demonstrated with my assignments would tend to confirm my hunch. I expect that in this golden age of information having such a tendency would probably result in very extreme achievement in those with such a trait.

    It should be noted that the informational resources that I can access through my personal computer from my online university are considerable. When I enroll in a course there are over 30,000 reference journals that can be found in the virtual library. It is quite impressive. This could be provided to students anywhere in the world who had internet access without respect to the per capita GDP of their nation etc. It would be a fantastic way to leapfrog past the enormous expense that has been required to build up the cognitive resource base of developed nations. I am not sure how many physical university libraries would have such an extensive collection. I can only hope that similar access is now offered to all students at any level in their education. It makes a tremendous difference when you can access the same material that might be written from even a slightly different perspective. I have noticed that if you only read one account of a subject that you will often miss out important aspects of the subject material.

    What I find especially amusing about this is that this is one of the few questions in psychometrics that could easily be demonstrated in an experimental setting any time into the future, no matter how elaborate infotech becomes. All you would need to do is put a test subject in a environment for say a day and ask them to answer some research question: in one setting (control) they would only have access to say a high school library, in another setting (treatment) they would have access to a computer would enhanced internet access (i.e., research journals , ebooks). It should be obvious that with internet access the measured performance of the work would be enormously higher than with only a library.

    It is surprising to me that more has not been said about this. It is as if we superturbocharged the cognitive space and no one thought this was worth commenting on. While those on this blog are highly psychometrically aware and will likely call me on this, but I have to say this new internet environment has probably added about 30 IQ points to the quality of assignment that I can now submit. Up to this point there has not been a single question on any of my assignments that I could not research endlessly. In one of my assignments I was able to simply google the question and a quote on this exact question popped up from a Nobel winning scientist. Without the internet there would have been no possible way that I could have found this quote out of the tens of millions of books in print. The internet has quite clearly logarithmically amplified my intellectual reach.

    new internet environment has probably added about 30 IQ points to the quality of assignment that I can now submit.

    This is an interesting point (though I am not sure the unit of measurement is appropriate). The areas where I see the biggest improvements are:
    - Professional presentation. Formatting, adherence to guidelines (like your APA style comments above), spell checking.
    - Reference availability and searchability as you have discussed.
    - Data and analytics. And graphical presentation of data.
    - Source control. Being able to keep good records of changes to data, analytics, and presentation is incredibly helpful.
    - Replicability. I can hand someone an R markdown file and my data and they can reproduce (and extend) my results exactly. In a well formatted fashion.

    I find the combination of R, RStudio, R Markdown, and https://rpubs.com/ incredibly powerful as an environment for accomplishing all of the above. Davide Piffer’s work provides a good example of what is possible. Both with RPubs: http://rpubs.com/Daxide/ and OSF: https://osf.io/ewxqj/
    Emil’s website also has good examples: http://emilkirkegaard.dk/en/

    But none of those are more important than the ability to ask good questions, think clearly about them, draw useful conclusions, and express both the reasoning and the conclusions clearly and in a well supported fashion to others. The new tools just make doing so much easier–if you have the ability. Though one can argue the tools also make it easier to acquire the ability (the barriers to entry are lower).

    I wonder how all of this appears to professors critiquing student assignments. Are the improvements real, or are they just “turd polishing”? If you have any older professors (or if any here want to comment) perhaps you could ask them about this?

    there has not been a single question on any of my assignments that I could not research endlessly.

    The type of person who has a compulsion to do this as things interest them (me too) is exactly the type of person who thrives on this availability. I think many just find it overwhelming (as I do sometimes). I am not sure the improvement has been as large for the dedicated academic researcher (I would be interested in hearing counterpoints). For decades there have been libraries, books, journals, and photocopiers. And I think academics are (and have always been) quite good at accumulating the resources they need as individuals, labs, and fields.

    Without the internet there would have been no possible way that I could have found this quote out of the tens of millions of books in print.

    It is daunting to think how much harder things like that used to be. Even with access to a high end reference library. The ability we have now to find information and then track it back through a chain of references is mind boggling. Then there is the ability to link directly to those references either in whole or part (I find selected figures particular informative, not sure how others feel about that ; ).

    It is important to note that it is not just “the internet.” It is the contents as well as the connectivity. Years ago it seemed like “everything is available” until you actually went and looked. Now almost everything really is available electronically even if not all of it is public. Getting all those books and papers digitized has taken time.

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

    res, I think this is a big one! I do not think it has been as fully explored nearly as much as it should have been. From what I can see from my old assignments there has clearly been a large shift in the quality of my submissions. I find it quite startlingly to realize that some people would likely have been many years ahead of me on the technology curve and this would have given them a truly overwhelming edge. Surprisingly some of my tutors have mentioned that some people are still submitting non-computer aided assignments. This is almost beyond my comprehension. The difference in the quality of assignments with and without the range of available technology (which as you noted extends beyond simply the internet as the world’s biggest research library) would simply be massive. I would have to think it would represent an effective cognitive distance of up to 10 grade levels. Some of my recent efforts I would self describe as brilliant. They work on a variety of levels and most importantly they actually made contributions to the fields in moving the discussion forward.

    The 30 IQ point uplift to my assignment quality statement could be thought of in terms of the assignments being a Turing IQ test. I would be interested to know what professional psychometricians might evaluate my IQ to be if they could only view assignments from either the pre or post internet era. When I reviewed the assignments today I simply cringed while looking at them. Most of the assignments had few or any references. Writing an assignment based primarily on your own observations really is not considered to be scholarship at all. It had never been obvious to me where I was supposed to go to find such references There is almost no chance that our school library would have had the fresh and relevant resources that I would have needed. Yet, today I can type in almost any research topic into my web browser and I could find an ocean of information on anything.

    Yes, you are quite right about the professional presentation side of things. My current assignments have very high production value. Several of my tutors have made favorable comments of my use of figures in my assignments. No one wants to fight through a wall of words. I try and put in a figure break every 2-3 pages to make my assignments fun to read. I can use my speller and grammar checker, and also create nice and clean formatting. I have asked my online University to consider upgrading the software provided so that APA citation spelling, plagiarism, possibly virtual writing coaches etc. could all be provided to students. I don’t understand why this policy has it been adopted by all the major journals already. In a current assignment that I am working on it turns out that a published article did not use proper APA style in it’s title. I was able to find an article that cited this improper title in it’s referencing. Strangely, the citing article also had a mistake in it’s APA title format. All these low level style questions, spelling , and grammar questions distract from actually focusing on important substantive questions of the content of the articles. It is very surprising to me that the leadership in the scientific community has not stepped forward and imposed a computerized algorithm that would force consistent application of formatting and other rules.

    I have also found it very helpful that the code for articles such as for the article of this thread have been provided. I am very unsure how articles that do not provide such code or extensive documentation of the output can actually be regarded as science. I was wondering lately how much data siloing is going on out there. It occurred to me that there should be no particularly coherent reason why a dataset could not be remixed multiple times using a range of assumptions and methods. Yet, this does not often seem to be what happens. A dataset typically will only have one group as the Monopoly owner of the data. Considering how much money is now being invested into acquiring these GWAS genotypes, I would like to know what an open competitive process might be able to reveal.

    With the current article, I had quite a bit of fun forking the R code, even though I have little understanding of R. All I had to do was take the code that was given and remix it. That sort of an open process is what science should be all about.

    I think a major point here is that the information resources that are available to you greatly open up the frontiers of what topics that you can explore. A perfect example of this is from my previous assignment. There was a list of suggested topics for the assignment and I read through the list and I was not interested in any of the topics. In previous courses that I have taken there was no choice. YOU had to select a topic from the list. I was very happy when they oked my idea of going off road with the psychometric idea.

    It is true that a substantial amount of resources do exist and have existed for probably centuries in academic libraries, though for me having the right skill set (social, research etc.) was lacking. Even if I were given all the resources that I can now access online in a paper format in a physical library I would still have an enormous advantage with virtual technology. The barriers that existed are no longer are present. I can access a massive research library from my own home 24/7. As a guess my access of academic literature has increased by a factor of 1000 when comparing pre versus post internet.

    That is also a good point that this is still ramping up. We are still struggling with the idea of a truly open digital space. I respect and understand the need for copyright, though I think that we have probably entered a time in which scientific and other idea works likely should only have about 20 year effective protection. A large number of journals have moved to only a few year protection, though perhaps the entire library of human knowledge should open up before it has lost it’s freshness.

    Read More
    • Replies: @res

    With the current article, I had quite a bit of fun forking the R code, even though I have little understanding of R. All I had to do was take the code that was given and remix it.
     
    I did the same thing. One thing I did was calculate -log10(p) (a standard measure of significance in the GWAS literature) for the SNPs which makes it much easier to look at things like histograms of SNP significance by p-value. Another thing was to look at SNPs in the exclusive intervals with p < 10^-30 and 10^-30 < p < 10^-20. I was surprised to find there was almost 0 (|| < 0.06 for both populations and continents) correlation between the PGS computed from those subsets. I think this calls into question the "small number of SNPs show selection pressure so conclusions are applicable to the other SNPs" idea. Though perhaps the more limited version in Piffer's work is still applicable. Namely:
    1. Very small number of SNPs (e.g. 9) which would most clearly show selection pressure.
    2. Limiting to SNPs with positive effect for the derived allele.

    The subtlety of 2. is something I only really noticed recently in Piffer's early work. I initially worried about it as cherry picking, but I am coming to think of it as sensible. The interesting question there is what is happening with the SNPs that have a negative effect for the derived allele yet have managed to achieve a somewhat high frequency. IIRC those are the SNPs where the lower IQ variant is non-intuitively more frequent in higher IQ populations.

    It is important to note that the low correlation is seen only with exclusive subsets. PGS from p < 10^-30 and p < 10^-20 correlate at 0.77. This is what I was alluding to in my A + B statement in comment 85.

    To draw more firm conclusions it would be good to look at the SNP characteristics more closely by subset. Are the derived alleles typically + or -, how much (and with what pattern) do the frequencies vary between populations, etc.?

    If you want to learn R, how to use it effectively, and see the breadth of tools and capabilities (e.g. RStudio, R Markdown, RPubs, Shiny interactive web apps, GitHub) available I highly recommend this series of courses: https://www.coursera.org/specializations/jhu-data-science

    It looks like they are no longer free, but if you are a tuition paying student I think you will see they offer a great deal of value for the money ($49 per class).

    I think course 5, Reproducible Research, is especially relevant to the conversation we are having now: https://www.coursera.org/learn/reproducible-research

    P.S. Some fairly simple code you might find helpful.

    gwas$log10p 7)], breaks=7:ceiling(max(gwas$log10p))) # Roughly significant SNPs
     

    # Use ranges instead
    hpthresh <- 10^-20
    lpthresh <- 10^-30
    gwas=gwas[which(gwas$Pvallpthresh),] # select SNPs within range
    subset.note <- paste(lpthresh, "< p <", hpthresh)

    hist(gwas$log10p, breaks=max(gwas$log10p), main=paste("SNP Frequency Distribution for", subset.note))

    cat("Using SNP subset", subset.note)
     
    I then output the PGS results for both populations and continents so I could load them as below. There is an annoying issue where R Markdown uses UTF-8 style double quotes which I had to convert to " below.

    ```{r, results='asis'}
    cat("Using SNP subset", subset.note)
    # Formatted data for cut and paste from knitted HTML
    dump("df_PGS", "")
    ```

    If you want to check my work, here is how I generated my results:

    # Using SNP subset p < 5e-08
    df_PGS <- structure(list(PGS_Z = c(0.5200636075001, -1.11902371997781, -1.02208247762685, -2.07651476702638, -0.302255484493313, -1.42211430460949, 1.40980466840254, 0.559582471229181, 1.09486218318298, 0.504133511124049, -0.466988474444203, 0.967393318105655, 1.69173244649728, 1.12514986737479, 1.67521351362267, 0.326979905514463, -0.310754319990793, 1.32151715857769, 0.0685786322416409, 1.14807956305697, 0.788887780341753, -0.372752615000197, 0.860232668360736, 1.3287069623234, 0.92922372898383, -0.581182512307871, -0.224925800495491, -0.764091225900405, 0.185373269728186, -0.7909882596934, 0.369592154021238, 0.248155541120252, -0.530569528552552, -1.28424949508351, -0.946197739930438, -1.72471514128299, -3.18629540939235, -1.19153830835272, 1.02634720313318, 0.338840917692266, -0.222976299902283, 0.568804159526946, 0.68520661899737, 0.40549941129595, -0.32654166812477, -0.594762957433412, 0.0515190783651431, -0.0106906814603098, 0.308741160042542, -0.246360815781283, -0.787651899184517, 0.755315367983932, -0.757312962299404), Population = c("Algeria (Mzab) - Mozabite", "Bougainville - NAN Melanesian", "Brazil - Karitiana", "Brazil - Surui", "C. African Republic - Biaka Pygmy", "Cambodia - Cambodian", "China - Dai", "China - Daur", "China - Han", "China - Hezhen", "China - Lahu", "China - Miaozu", "China - Mongola", "China - Naxi", "China - Oroqen", "China - She", "China - Tu", "China - Tujia", "China - Uygur", "China - Xibo", "China - Yizu", "Colombia - Piapoco and Curripaco", "D. R. of Congo - Mbuti Pygmy", "France - Basque", "France - French", "Israel (Carmel) - Druze", "Israel (Central) - Palestinian", "Israel (Negev) - Bedouin", "Italy - from Bergamo", "Italy - Sardinian", "Italy - Tuscan", "Japan - Japanese", "Kenya - Bantu", "Mexico - Maya", "Mexico - Pima", "Namibia - San", "New Guinea - Papuan", "Nigeria - Yoruba", "Orkney Islands - Orcadian", "Pakistan - Balochi", "Pakistan - Brahui", "Pakistan - Burusho", "Pakistan - Hazara", "Pakistan - Kalash", "Pakistan - Makrani", "Pakistan - Pathan", "Pakistan - Sindhi", "Population Set 1", "Russia - Russian", "Russia (Caucasus) - Adygei", "Senegal - Mandenka", "Siberia - Yakut", "South Africa - Bantu" )), .Names = c("PGS_Z", "Population"), row.names = c("Algeria (Mzab) - Mozabite", "Bougainville - NAN Melanesian", "Brazil - Karitiana", "Brazil - Surui", "C. African Republic - Biaka Pygmy", "Cambodia - Cambodian", "China - Dai", "China - Daur", "China - Han", "China - Hezhen", "China - Lahu", "China - Miaozu", "China - Mongola", "China - Naxi", "China - Oroqen", "China - She", "China - Tu", "China - Tujia", "China - Uygur", "China - Xibo", "China - Yizu", "Colombia - Piapoco and Curripaco", "D. R. of Congo - Mbuti Pygmy", "France - Basque", "France - French", "Israel (Carmel) - Druze", "Israel (Central) - Palestinian", "Israel (Negev) - Bedouin", "Italy - from Bergamo", "Italy - Sardinian", "Italy - Tuscan", "Japan - Japanese", "Kenya - Bantu", "Mexico - Maya", "Mexico - Pima", "Namibia - San", "New Guinea - Papuan", "Nigeria - Yoruba", "Orkney Islands - Orcadian", "Pakistan - Balochi", "Pakistan - Brahui", "Pakistan - Burusho", "Pakistan - Hazara", "Pakistan - Kalash", "Pakistan - Makrani", "Pakistan - Pathan", "Pakistan - Sindhi", "Population Set 1", "Russia - Russian", "Russia (Caucasus) - Adygei", "Senegal - Mandenka", "Siberia - Yakut", "South Africa - Bantu" ), class = "data.frame")
    df_PGS <- structure(list(PGS_Z = c(-0.171877416500972, -0.76325883175675, 0.527052463582658, 1.13366534856215, 0.802379817204429, 0.0721060622819543, -2.00331924902356, 0.403251805650091), Population = c("AFRICA", "AMERICA", "CENTRAL-SOUTH ASIA", "EAST ASIA", "EUROPE", "MIDDLE EAST", "OCEANIA", "Population Set 1")), .Names = c("PGS_Z", "Population" ), row.names = c("AFRICA", "AMERICA", "CENTRAL-SOUTH ASIA", "EAST ASIA", "EUROPE", "MIDDLE EAST", "OCEANIA", "Population Set 1"), class = "data.frame")

    # Using SNP subset p < 1e-30
    df_PGS_pop_30 <- structure(list(PGS_Z = c(-0.22774097619992, -1.9715591510291, -0.671060706737492, 0.640248446937808, 0.498481462070653, -0.165132021585292, 1.20605525284874, 1.59005582149218, 1.5507521144634, 1.18022912859246, 0.38682595275503, 1.01660713020854, 0.622694079610597, 1.13644957931209, 0.808396343591038, 0.188319081351515, 1.12763415841273, 0.230240996196302, 0.816383059308068, 0.374866735853536, 0.656620936068732, 0.201996888194904, -0.421196361148007, -0.829255798533044, -0.637466166619254, -0.431621917284199, -0.86857201956625, -1.08538270811792, -1.77791883690442, -2.02167495788651, -0.437408949111198, 2.02081705336049, 0.22021588820191, -1.40166582744889, -0.650943749391742, -0.549937657847137, -2.65369336124334, 0.158268785377203, -0.43707246143649, 0.0229409510124096, -0.216314299709225, 0.928077501081921, 1.12058182169323, 1.20002072182299, -1.0069769085762, -0.249006441063474, -0.237998288498758, -0.0907098467875349, -1.0558649525544, -0.795996355644576, 0.335956524741878, 1.11970862227292, -0.467274315908915), Population = c("Algeria (Mzab) - Mozabite", "Bougainville - NAN Melanesian", "Brazil - Karitiana", "Brazil - Surui", "C. African Republic - Biaka Pygmy", "Cambodia - Cambodian", "China - Dai", "China - Daur", "China - Han", "China - Hezhen", "China - Lahu", "China - Miaozu", "China - Mongola", "China - Naxi", "China - Oroqen", "China - She", "China - Tu", "China - Tujia", "China - Uygur", "China - Xibo", "China - Yizu", "Colombia - Piapoco and Curripaco", "D. R. of Congo - Mbuti Pygmy", "France - Basque", "France - French", "Israel (Carmel) - Druze", "Israel (Central) - Palestinian", "Israel (Negev) - Bedouin", "Italy - from Bergamo", "Italy - Sardinian", "Italy - Tuscan", "Japan - Japanese", "Kenya - Bantu", "Mexico - Maya", "Mexico - Pima", "Namibia - San", "New Guinea - Papuan", "Nigeria - Yoruba", "Orkney Islands - Orcadian", "Pakistan - Balochi", "Pakistan - Brahui", "Pakistan - Burusho", "Pakistan - Hazara", "Pakistan - Kalash", "Pakistan - Makrani", "Pakistan - Pathan", "Pakistan - Sindhi", "Population Set 1", "Russia - Russian", "Russia (Caucasus) - Adygei", "Senegal - Mandenka", "Siberia - Yakut", "South Africa - Bantu" )), .Names = c("PGS_Z", "Population"), row.names = c("Algeria (Mzab) - Mozabite", "Bougainville - NAN Melanesian", "Brazil - Karitiana", "Brazil - Surui", "C. African Republic - Biaka Pygmy", "Cambodia - Cambodian", "China - Dai", "China - Daur", "China - Han", "China - Hezhen", "China - Lahu", "China - Miaozu", "China - Mongola", "China - Naxi", "China - Oroqen", "China - She", "China - Tu", "China - Tujia", "China - Uygur", "China - Xibo", "China - Yizu", "Colombia - Piapoco and Curripaco", "D. R. of Congo - Mbuti Pygmy", "France - Basque", "France - French", "Israel (Carmel) - Druze", "Israel (Central) - Palestinian", "Israel (Negev) - Bedouin", "Italy - from Bergamo", "Italy - Sardinian", "Italy - Tuscan", "Japan - Japanese", "Kenya - Bantu", "Mexico - Maya", "Mexico - Pima", "Namibia - San", "New Guinea - Papuan", "Nigeria - Yoruba", "Orkney Islands - Orcadian", "Pakistan - Balochi", "Pakistan - Brahui", "Pakistan - Burusho", "Pakistan - Hazara", "Pakistan - Kalash", "Pakistan - Makrani", "Pakistan - Pathan", "Pakistan - Sindhi", "Population Set 1", "Russia - Russian", "Russia (Caucasus) - Adygei", "Senegal - Mandenka", "Siberia - Yakut", "South Africa - Bantu" ), class = "data.frame")
    df_PGS_cont_30 <- structure(list(PGS_Z = c(0.534009086642049, -0.207818862875531, 0.624192430205281, 1.47355288841214, -0.608538124342132, -0.258024397998142, -1.88798149112997, 0.330608471086306), Population = c("AFRICA", "AMERICA", "CENTRAL-SOUTH ASIA", "EAST ASIA", "EUROPE", "MIDDLE EAST", "OCEANIA", "Population Set 1")), .Names = c("PGS_Z", "Population" ), row.names = c("AFRICA", "AMERICA", "CENTRAL-SOUTH ASIA", "EAST ASIA", "EUROPE", "MIDDLE EAST", "OCEANIA", "Population Set 1"), class = "data.frame")

    # Using SNP subset p < 1e-20
    df_PGS_pop_20 <- structure(list(PGS_Z = c(0.786289242132707, -1.53926733341709, -0.282415142753057, -0.861436091656981, 0.715737084448114, -0.621032150920052, 1.01071195358365, 0.676893455546647, 1.36740374951033, 0.672415354678118, 1.40774807076906, 1.23085709241133, 1.54303815685287, 1.61726957190717, 1.36121243993766, 0.418990155927757, 1.10541795637635, 0.711390852697397, 0.408418829954349, 0.703152847230079, 0.704742905435769, -0.183833713655054, 1.34003273385856, 0.210120371586764, 0.222845196541972, -0.813324570376352, -0.412441362398562, -1.27382024839739, -0.859352341630058, -1.07915854413014, -0.0302683115662864, 1.76819650181227, 0.0458070914825576, -1.98327828188413, -1.2368391345566, -0.273867626298476, -2.35709998519742, -0.940251320839552, 0.0556024603329693, -0.439361189498476, -0.758666455784151, -0.197579706902746, 0.618366453613662, -0.234761348988804, -1.51306679241434, -1.05135377716523, -0.781881959483692, -0.140531600778617, -1.26428098899066, -0.885158365106794, -0.0578638316665617, 0.906023131624307, 0.463508516204846), Population = c("Algeria (Mzab) - Mozabite", "Bougainville - NAN Melanesian", "Brazil - Karitiana", "Brazil - Surui", "C. African Republic - Biaka Pygmy", "Cambodia - Cambodian", "China - Dai", "China - Daur", "China - Han", "China - Hezhen", "China - Lahu", "China - Miaozu", "China - Mongola", "China - Naxi", "China - Oroqen", "China - She", "China - Tu", "China - Tujia", "China - Uygur", "China - Xibo", "China - Yizu", "Colombia - Piapoco and Curripaco", "D. R. of Congo - Mbuti Pygmy", "France - Basque", "France - French", "Israel (Carmel) - Druze", "Israel (Central) - Palestinian", "Israel (Negev) - Bedouin", "Italy - from Bergamo", "Italy - Sardinian", "Italy - Tuscan", "Japan - Japanese", "Kenya - Bantu", "Mexico - Maya", "Mexico - Pima", "Namibia - San", "New Guinea - Papuan", "Nigeria - Yoruba", "Orkney Islands - Orcadian", "Pakistan - Balochi", "Pakistan - Brahui", "Pakistan - Burusho", "Pakistan - Hazara", "Pakistan - Kalash", "Pakistan - Makrani", "Pakistan - Pathan", "Pakistan - Sindhi", "Population Set 1", "Russia - Russian", "Russia (Caucasus) - Adygei", "Senegal - Mandenka", "Siberia - Yakut", "South Africa - Bantu" )), .Names = c("PGS_Z", "Population"), row.names = c("Algeria (Mzab) - Mozabite", "Bougainville - NAN Melanesian", "Brazil - Karitiana", "Brazil - Surui", "C. African Republic - Biaka Pygmy", "Cambodia - Cambodian", "China - Dai", "China - Daur", "China - Han", "China - Hezhen", "China - Lahu", "China - Miaozu", "China - Mongola", "China - Naxi", "China - Oroqen", "China - She", "China - Tu", "China - Tujia", "China - Uygur", "China - Xibo", "China - Yizu", "Colombia - Piapoco and Curripaco", "D. R. of Congo - Mbuti Pygmy", "France - Basque", "France - French", "Israel (Carmel) - Druze", "Israel (Central) - Palestinian", "Israel (Negev) - Bedouin", "Italy - from Bergamo", "Italy - Sardinian", "Italy - Tuscan", "Japan - Japanese", "Kenya - Bantu", "Mexico - Maya", "Mexico - Pima", "Namibia - San", "New Guinea - Papuan", "Nigeria - Yoruba", "Orkney Islands - Orcadian", "Pakistan - Balochi", "Pakistan - Brahui", "Pakistan - Burusho", "Pakistan - Hazara", "Pakistan - Kalash", "Pakistan - Makrani", "Pakistan - Pathan", "Pakistan - Sindhi", "Population Set 1", "Russia - Russian", "Russia (Caucasus) - Adygei", "Senegal - Mandenka", "Siberia - Yakut", "South Africa - Bantu" ), class = "data.frame")
    df_PGS_cont_20 <- structure(list(PGS_Z = c(0.653893673462433, -0.730733849600686, -0.0670803805416632, 1.69027530476756, -0.0376922725492349, -0.102939430156396, -1.74614356692521, 0.3404205215432), Population = c("AFRICA", "AMERICA", "CENTRAL-SOUTH ASIA", "EAST ASIA", "EUROPE", "MIDDLE EAST", "OCEANIA", "Population Set 1")), .Names = c("PGS_Z", "Population" ), row.names = c("AFRICA", "AMERICA", "CENTRAL-SOUTH ASIA", "EAST ASIA", "EUROPE", "MIDDLE EAST", "OCEANIA", "Population Set 1"), class = "data.frame")

    # Using SNP subset 1e-30 < p < 1e-20
    df_PGS_pop_20_30 <- structure(list(PGS_Z = c(1.50332824545052, 0.00940525814543819, 0.379513465320102, -2.12421853700429, 0.506455789880328, -0.766008854773309, 0.101919149222941, -0.887214805243339, 0.236732431127176, -0.393669269400149, 1.72110498389014, 0.676320728858494, 1.64384563646339, 1.13206019709658, 1.13372778719622, 0.422870807910656, 0.345265036471351, 0.82728687942642, -0.360658796949194, 0.637818416914439, 0.296188301359843, -0.533147388375364, 2.60243314417584, 1.34018121526478, 1.12577531831984, -0.740171397170485, 0.418129553778085, -0.659269503141165, 0.832347387060755, 0.787548309241492, 0.48704585623588, 0.28715071020886, -0.197574598569653, -1.37846103661106, -1.13222828259139, 0.24490396224678, -0.431838742587047, -1.6583835258514, 0.620437780794498, -0.712625593701843, -0.917959936357313, -1.44133190041844, -0.405040128165086, -1.83139345145332, -1.12781816905905, -1.33409498880749, -0.927651317924142, -0.108190410703586, -0.680455708928403, -0.407089803572685, -0.500468056524642, 0.0442497400746086, 1.2929181117484), Population = c("Algeria (Mzab) - Mozabite", "Bougainville - NAN Melanesian", "Brazil - Karitiana", "Brazil - Surui", "C. African Republic - Biaka Pygmy", "Cambodia - Cambodian", "China - Dai", "China - Daur", "China - Han", "China - Hezhen", "China - Lahu", "China - Miaozu", "China - Mongola", "China - Naxi", "China - Oroqen", "China - She", "China - Tu", "China - Tujia", "China - Uygur", "China - Xibo", "China - Yizu", "Colombia - Piapoco and Curripaco", "D. R. of Congo - Mbuti Pygmy", "France - Basque", "France - French", "Israel (Carmel) - Druze", "Israel (Central) - Palestinian", "Israel (Negev) - Bedouin", "Italy - from Bergamo", "Italy - Sardinian", "Italy - Tuscan", "Japan - Japanese", "Kenya - Bantu", "Mexico - Maya", "Mexico - Pima", "Namibia - San", "New Guinea - Papuan", "Nigeria - Yoruba", "Orkney Islands - Orcadian", "Pakistan - Balochi", "Pakistan - Brahui", "Pakistan - Burusho", "Pakistan - Hazara", "Pakistan - Kalash", "Pakistan - Makrani", "Pakistan - Pathan", "Pakistan - Sindhi", "Population Set 1", "Russia - Russian", "Russia (Caucasus) - Adygei", "Senegal - Mandenka", "Siberia - Yakut", "South Africa - Bantu" )), .Names = c("PGS_Z", "Population"), row.names = c("Algeria (Mzab) - Mozabite", "Bougainville - NAN Melanesian", "Brazil - Karitiana", "Brazil - Surui", "C. African Republic - Biaka Pygmy", "Cambodia - Cambodian", "China - Dai", "China - Daur", "China - Han", "China - Hezhen", "China - Lahu", "China - Miaozu", "China - Mongola", "China - Naxi", "China - Oroqen", "China - She", "China - Tu", "China - Tujia", "China - Uygur", "China - Xibo", "China - Yizu", "Colombia - Piapoco and Curripaco", "D. R. of Congo - Mbuti Pygmy", "France - Basque", "France - French", "Israel (Carmel) - Druze", "Israel (Central) - Palestinian", "Israel (Negev) - Bedouin", "Italy - from Bergamo", "Italy - Sardinian", "Italy - Tuscan", "Japan - Japanese", "Kenya - Bantu", "Mexico - Maya", "Mexico - Pima", "Namibia - San", "New Guinea - Papuan", "Nigeria - Yoruba", "Orkney Islands - Orcadian", "Pakistan - Balochi", "Pakistan - Brahui", "Pakistan - Burusho", "Pakistan - Hazara", "Pakistan - Kalash", "Pakistan - Makrani", "Pakistan - Pathan", "Pakistan - Sindhi", "Population Set 1", "Russia - Russian", "Russia (Caucasus) - Adygei", "Senegal - Mandenka", "Siberia - Yakut", "South Africa - Bantu" ), class = "data.frame")
    df_PGS_cont_20_30 <- structure(list(PGS_Z = c(0.438789045972685, -1.35288548955514, -1.55073254279487, 0.927621371934068, 1.25596106648565, 0.316698910098827, -0.147239425870863, 0.11178706372964), Population = c("AFRICA", "AMERICA", "CENTRAL-SOUTH ASIA", "EAST ASIA", "EUROPE", "MIDDLE EAST", "OCEANIA", "Population Set 1")), .Names = c("PGS_Z", "Population" ), row.names = c("AFRICA", "AMERICA", "CENTRAL-SOUTH ASIA", "EAST ASIA", "EUROPE", "MIDDLE EAST", "OCEANIA", "Population Set 1"), class = "data.frame")

    library(corrplot)
    cor(df_PGS_pop_30$PGS_Z, df_PGS_pop_20_30$PGS_Z)
    corrplot(cor(data.frame(df_PGS_pop_20$PGS_Z, df_PGS_pop_30$PGS_Z, df_PGS_pop_20_30$PGS_Z)))

    cor(df_PGS_cont_30$PGS_Z, df_PGS_cont_20_30$PGS_Z)
    corrplot(cor(data.frame(df_PGS_cont_20$PGS_Z, df_PGS_cont_30$PGS_Z, df_PGS_cont_20_30$PGS_Z)))

     

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

    res, I was also excited about how academic work could be used as a measure of National achievement.
    Instead of focusing in on teaching to standardized tests such as SAT or PISA, one could have students submit their academic work online (probably best to do this anonymously). This could offer a tremendous insight into the psychometric potential of the entire world. With standardized tests, there is a legitimate question whether having all the students working on the same generic problems offers any marginal utility. Allowing students to research and publish online topics that they were interested in could result in substantial contributions to the knowledge base. At this time most of this youthful energy is not captured.

    In addition, having an online repository of academic output could be a great way to document the psychometric changes that occur as technology has advanced.

    Read More
    • Replies: @res
    That is an interesting idea. It also offers great potential as a talent search (e.g. though anonymous, allow email contact through an intermediary).
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  95. res says:
    @Factorize
    res, I think this is a big one! I do not think it has been as fully explored nearly as much as it should have been. From what I can see from my old assignments there has clearly been a large shift in the quality of my submissions. I find it quite startlingly to realize that some people would likely have been many years ahead of me on the technology curve and this would have given them a truly overwhelming edge. Surprisingly some of my tutors have mentioned that some people are still submitting non-computer aided assignments. This is almost beyond my comprehension. The difference in the quality of assignments with and without the range of available technology (which as you noted extends beyond simply the internet as the world's biggest research library) would simply be massive. I would have to think it would represent an effective cognitive distance of up to 10 grade levels. Some of my recent efforts I would self describe as brilliant. They work on a variety of levels and most importantly they actually made contributions to the fields in moving the discussion forward.

    The 30 IQ point uplift to my assignment quality statement could be thought of in terms of the assignments being a Turing IQ test. I would be interested to know what professional psychometricians might evaluate my IQ to be if they could only view assignments from either the pre or post internet era. When I reviewed the assignments today I simply cringed while looking at them. Most of the assignments had few or any references. Writing an assignment based primarily on your own observations really is not considered to be scholarship at all. It had never been obvious to me where I was supposed to go to find such references There is almost no chance that our school library would have had the fresh and relevant resources that I would have needed. Yet, today I can type in almost any research topic into my web browser and I could find an ocean of information on anything.

    Yes, you are quite right about the professional presentation side of things. My current assignments have very high production value. Several of my tutors have made favorable comments of my use of figures in my assignments. No one wants to fight through a wall of words. I try and put in a figure break every 2-3 pages to make my assignments fun to read. I can use my speller and grammar checker, and also create nice and clean formatting. I have asked my online University to consider upgrading the software provided so that APA citation spelling, plagiarism, possibly virtual writing coaches etc. could all be provided to students. I don't understand why this policy has it been adopted by all the major journals already. In a current assignment that I am working on it turns out that a published article did not use proper APA style in it's title. I was able to find an article that cited this improper title in it's referencing. Strangely, the citing article also had a mistake in it's APA title format. All these low level style questions, spelling , and grammar questions distract from actually focusing on important substantive questions of the content of the articles. It is very surprising to me that the leadership in the scientific community has not stepped forward and imposed a computerized algorithm that would force consistent application of formatting and other rules.

    I have also found it very helpful that the code for articles such as for the article of this thread have been provided. I am very unsure how articles that do not provide such code or extensive documentation of the output can actually be regarded as science. I was wondering lately how much data siloing is going on out there. It occurred to me that there should be no particularly coherent reason why a dataset could not be remixed multiple times using a range of assumptions and methods. Yet, this does not often seem to be what happens. A dataset typically will only have one group as the Monopoly owner of the data. Considering how much money is now being invested into acquiring these GWAS genotypes, I would like to know what an open competitive process might be able to reveal.

    With the current article, I had quite a bit of fun forking the R code, even though I have little understanding of R. All I had to do was take the code that was given and remix it. That sort of an open process is what science should be all about.

    I think a major point here is that the information resources that are available to you greatly open up the frontiers of what topics that you can explore. A perfect example of this is from my previous assignment. There was a list of suggested topics for the assignment and I read through the list and I was not interested in any of the topics. In previous courses that I have taken there was no choice. YOU had to select a topic from the list. I was very happy when they oked my idea of going off road with the psychometric idea.

    It is true that a substantial amount of resources do exist and have existed for probably centuries in academic libraries, though for me having the right skill set (social, research etc.) was lacking. Even if I were given all the resources that I can now access online in a paper format in a physical library I would still have an enormous advantage with virtual technology. The barriers that existed are no longer are present. I can access a massive research library from my own home 24/7. As a guess my access of academic literature has increased by a factor of 1000 when comparing pre versus post internet.

    That is also a good point that this is still ramping up. We are still struggling with the idea of a truly open digital space. I respect and understand the need for copyright, though I think that we have probably entered a time in which scientific and other idea works likely should only have about 20 year effective protection. A large number of journals have moved to only a few year protection, though perhaps the entire library of human knowledge should open up before it has lost it's freshness.

    With the current article, I had quite a bit of fun forking the R code, even though I have little understanding of R. All I had to do was take the code that was given and remix it.

    I did the same thing. One thing I did was calculate -log10(p) (a standard measure of significance in the GWAS literature) for the SNPs which makes it much easier to look at things like histograms of SNP significance by p-value. Another thing was to look at SNPs in the exclusive intervals with p < 10^-30 and 10^-30 < p < 10^-20. I was surprised to find there was almost 0 (|| < 0.06 for both populations and continents) correlation between the PGS computed from those subsets. I think this calls into question the "small number of SNPs show selection pressure so conclusions are applicable to the other SNPs" idea. Though perhaps the more limited version in Piffer's work is still applicable. Namely:
    1. Very small number of SNPs (e.g. 9) which would most clearly show selection pressure.
    2. Limiting to SNPs with positive effect for the derived allele.

    The subtlety of 2. is something I only really noticed recently in Piffer's early work. I initially worried about it as cherry picking, but I am coming to think of it as sensible. The interesting question there is what is happening with the SNPs that have a negative effect for the derived allele yet have managed to achieve a somewhat high frequency. IIRC those are the SNPs where the lower IQ variant is non-intuitively more frequent in higher IQ populations.

    It is important to note that the low correlation is seen only with exclusive subsets. PGS from p < 10^-30 and p < 10^-20 correlate at 0.77. This is what I was alluding to in my A + B statement in comment 85.

    To draw more firm conclusions it would be good to look at the SNP characteristics more closely by subset. Are the derived alleles typically + or -, how much (and with what pattern) do the frequencies vary between populations, etc.?

    If you want to learn R, how to use it effectively, and see the breadth of tools and capabilities (e.g. RStudio, R Markdown, RPubs, Shiny interactive web apps, GitHub) available I highly recommend this series of courses: https://www.coursera.org/specializations/jhu-data-science

    It looks like they are no longer free, but if you are a tuition paying student I think you will see they offer a great deal of value for the money ($49 per class).

    I think course 5, Reproducible Research, is especially relevant to the conversation we are having now: https://www.coursera.org/learn/reproducible-research

    P.S. Some fairly simple code you might find helpful.

    [MORE]

    gwas$log10p 7)], breaks=7:ceiling(max(gwas$log10p))) # Roughly significant SNPs

    # Use ranges instead
    hpthresh <- 10^-20
    lpthresh <- 10^-30
    gwas=gwas[which(gwas$Pvallpthresh),] # select SNPs within range
    subset.note <- paste(lpthresh, "< p <", hpthresh)

    hist(gwas$log10p, breaks=max(gwas$log10p), main=paste("SNP Frequency Distribution for", subset.note))

    cat("Using SNP subset", subset.note)

    I then output the PGS results for both populations and continents so I could load them as below. There is an annoying issue where R Markdown uses UTF-8 style double quotes which I had to convert to ” below.

    “`{r, results=’asis’}
    cat(“Using SNP subset”, subset.note)
    # Formatted data for cut and paste from knitted HTML
    dump(“df_PGS”, “”)
    “`

    If you want to check my work, here is how I generated my results:

    # Using SNP subset p < 5e-08
    df_PGS <- structure(list(PGS_Z = c(0.5200636075001, -1.11902371997781, -1.02208247762685, -2.07651476702638, -0.302255484493313, -1.42211430460949, 1.40980466840254, 0.559582471229181, 1.09486218318298, 0.504133511124049, -0.466988474444203, 0.967393318105655, 1.69173244649728, 1.12514986737479, 1.67521351362267, 0.326979905514463, -0.310754319990793, 1.32151715857769, 0.0685786322416409, 1.14807956305697, 0.788887780341753, -0.372752615000197, 0.860232668360736, 1.3287069623234, 0.92922372898383, -0.581182512307871, -0.224925800495491, -0.764091225900405, 0.185373269728186, -0.7909882596934, 0.369592154021238, 0.248155541120252, -0.530569528552552, -1.28424949508351, -0.946197739930438, -1.72471514128299, -3.18629540939235, -1.19153830835272, 1.02634720313318, 0.338840917692266, -0.222976299902283, 0.568804159526946, 0.68520661899737, 0.40549941129595, -0.32654166812477, -0.594762957433412, 0.0515190783651431, -0.0106906814603098, 0.308741160042542, -0.246360815781283, -0.787651899184517, 0.755315367983932, -0.757312962299404), Population = c("Algeria (Mzab) – Mozabite", "Bougainville – NAN Melanesian", "Brazil – Karitiana", "Brazil – Surui", "C. African Republic – Biaka Pygmy", "Cambodia – Cambodian", "China – Dai", "China – Daur", "China – Han", "China – Hezhen", "China – Lahu", "China – Miaozu", "China – Mongola", "China – Naxi", "China – Oroqen", "China – She", "China – Tu", "China – Tujia", "China – Uygur", "China – Xibo", "China – Yizu", "Colombia – Piapoco and Curripaco", "D. R. of Congo – Mbuti Pygmy", "France – Basque", "France – French", "Israel (Carmel) – Druze", "Israel (Central) – Palestinian", "Israel (Negev) – Bedouin", "Italy – from Bergamo", "Italy – Sardinian", "Italy – Tuscan", "Japan – Japanese", "Kenya – Bantu", "Mexico – Maya", "Mexico – Pima", "Namibia – San", "New Guinea – Papuan", "Nigeria – Yoruba", "Orkney Islands – Orcadian", "Pakistan – Balochi", "Pakistan – Brahui", "Pakistan – Burusho", "Pakistan – Hazara", "Pakistan – Kalash", "Pakistan – Makrani", "Pakistan – Pathan", "Pakistan – Sindhi", "Population Set 1", "Russia – Russian", "Russia (Caucasus) – Adygei", "Senegal – Mandenka", "Siberia – Yakut", "South Africa – Bantu" )), .Names = c("PGS_Z", "Population"), row.names = c("Algeria (Mzab) – Mozabite", "Bougainville – NAN Melanesian", "Brazil – Karitiana", "Brazil – Surui", "C. African Republic – Biaka Pygmy", "Cambodia – Cambodian", "China – Dai", "China – Daur", "China – Han", "China – Hezhen", "China – Lahu", "China – Miaozu", "China – Mongola", "China – Naxi", "China – Oroqen", "China – She", "China – Tu", "China – Tujia", "China – Uygur", "China – Xibo", "China – Yizu", "Colombia – Piapoco and Curripaco", "D. R. of Congo – Mbuti Pygmy", "France – Basque", "France – French", "Israel (Carmel) – Druze", "Israel (Central) – Palestinian", "Israel (Negev) – Bedouin", "Italy – from Bergamo", "Italy – Sardinian", "Italy – Tuscan", "Japan – Japanese", "Kenya – Bantu", "Mexico – Maya", "Mexico – Pima", "Namibia – San", "New Guinea – Papuan", "Nigeria – Yoruba", "Orkney Islands – Orcadian", "Pakistan – Balochi", "Pakistan – Brahui", "Pakistan – Burusho", "Pakistan – Hazara", "Pakistan – Kalash", "Pakistan – Makrani", "Pakistan – Pathan", "Pakistan – Sindhi", "Population Set 1", "Russia – Russian", "Russia (Caucasus) – Adygei", "Senegal – Mandenka", "Siberia – Yakut", "South Africa – Bantu" ), class = "data.frame")
    df_PGS <- structure(list(PGS_Z = c(-0.171877416500972, -0.76325883175675, 0.527052463582658, 1.13366534856215, 0.802379817204429, 0.0721060622819543, -2.00331924902356, 0.403251805650091), Population = c("AFRICA", "AMERICA", "CENTRAL-SOUTH ASIA", "EAST ASIA", "EUROPE", "MIDDLE EAST", "OCEANIA", "Population Set 1")), .Names = c("PGS_Z", "Population" ), row.names = c("AFRICA", "AMERICA", "CENTRAL-SOUTH ASIA", "EAST ASIA", "EUROPE", "MIDDLE EAST", "OCEANIA", "Population Set 1"), class = "data.frame")

    # Using SNP subset p < 1e-30
    df_PGS_pop_30 <- structure(list(PGS_Z = c(-0.22774097619992, -1.9715591510291, -0.671060706737492, 0.640248446937808, 0.498481462070653, -0.165132021585292, 1.20605525284874, 1.59005582149218, 1.5507521144634, 1.18022912859246, 0.38682595275503, 1.01660713020854, 0.622694079610597, 1.13644957931209, 0.808396343591038, 0.188319081351515, 1.12763415841273, 0.230240996196302, 0.816383059308068, 0.374866735853536, 0.656620936068732, 0.201996888194904, -0.421196361148007, -0.829255798533044, -0.637466166619254, -0.431621917284199, -0.86857201956625, -1.08538270811792, -1.77791883690442, -2.02167495788651, -0.437408949111198, 2.02081705336049, 0.22021588820191, -1.40166582744889, -0.650943749391742, -0.549937657847137, -2.65369336124334, 0.158268785377203, -0.43707246143649, 0.0229409510124096, -0.216314299709225, 0.928077501081921, 1.12058182169323, 1.20002072182299, -1.0069769085762, -0.249006441063474, -0.237998288498758, -0.0907098467875349, -1.0558649525544, -0.795996355644576, 0.335956524741878, 1.11970862227292, -0.467274315908915), Population = c("Algeria (Mzab) – Mozabite", "Bougainville – NAN Melanesian", "Brazil – Karitiana", "Brazil – Surui", "C. African Republic – Biaka Pygmy", "Cambodia – Cambodian", "China – Dai", "China – Daur", "China – Han", "China – Hezhen", "China – Lahu", "China – Miaozu", "China – Mongola", "China – Naxi", "China – Oroqen", "China – She", "China – Tu", "China – Tujia", "China – Uygur", "China – Xibo", "China – Yizu", "Colombia – Piapoco and Curripaco", "D. R. of Congo – Mbuti Pygmy", "France – Basque", "France – French", "Israel (Carmel) – Druze", "Israel (Central) – Palestinian", "Israel (Negev) – Bedouin", "Italy – from Bergamo", "Italy – Sardinian", "Italy – Tuscan", "Japan – Japanese", "Kenya – Bantu", "Mexico – Maya", "Mexico – Pima", "Namibia – San", "New Guinea – Papuan", "Nigeria – Yoruba", "Orkney Islands – Orcadian", "Pakistan – Balochi", "Pakistan – Brahui", "Pakistan – Burusho", "Pakistan – Hazara", "Pakistan – Kalash", "Pakistan – Makrani", "Pakistan – Pathan", "Pakistan – Sindhi", "Population Set 1", "Russia – Russian", "Russia (Caucasus) – Adygei", "Senegal – Mandenka", "Siberia – Yakut", "South Africa – Bantu" )), .Names = c("PGS_Z", "Population"), row.names = c("Algeria (Mzab) – Mozabite", "Bougainville – NAN Melanesian", "Brazil – Karitiana", "Brazil – Surui", "C. African Republic – Biaka Pygmy", "Cambodia – Cambodian", "China – Dai", "China – Daur", "China – Han", "China – Hezhen", "China – Lahu", "China – Miaozu", "China – Mongola", "China – Naxi", "China – Oroqen", "China – She", "China – Tu", "China – Tujia", "China – Uygur", "China – Xibo", "China – Yizu", "Colombia – Piapoco and Curripaco", "D. R. of Congo – Mbuti Pygmy", "France – Basque", "France – French", "Israel (Carmel) – Druze", "Israel (Central) – Palestinian", "Israel (Negev) – Bedouin", "Italy – from Bergamo", "Italy – Sardinian", "Italy – Tuscan", "Japan – Japanese", "Kenya – Bantu", "Mexico – Maya", "Mexico – Pima", "Namibia – San", "New Guinea – Papuan", "Nigeria – Yoruba", "Orkney Islands – Orcadian", "Pakistan – Balochi", "Pakistan – Brahui", "Pakistan – Burusho", "Pakistan – Hazara", "Pakistan – Kalash", "Pakistan – Makrani", "Pakistan – Pathan", "Pakistan – Sindhi", "Population Set 1", "Russia – Russian", "Russia (Caucasus) – Adygei", "Senegal – Mandenka", "Siberia – Yakut", "South Africa – Bantu" ), class = "data.frame")
    df_PGS_cont_30 <- structure(list(PGS_Z = c(0.534009086642049, -0.207818862875531, 0.624192430205281, 1.47355288841214, -0.608538124342132, -0.258024397998142, -1.88798149112997, 0.330608471086306), Population = c("AFRICA", "AMERICA", "CENTRAL-SOUTH ASIA", "EAST ASIA", "EUROPE", "MIDDLE EAST", "OCEANIA", "Population Set 1")), .Names = c("PGS_Z", "Population" ), row.names = c("AFRICA", "AMERICA", "CENTRAL-SOUTH ASIA", "EAST ASIA", "EUROPE", "MIDDLE EAST", "OCEANIA", "Population Set 1"), class = "data.frame")

    # Using SNP subset p < 1e-20
    df_PGS_pop_20 <- structure(list(PGS_Z = c(0.786289242132707, -1.53926733341709, -0.282415142753057, -0.861436091656981, 0.715737084448114, -0.621032150920052, 1.01071195358365, 0.676893455546647, 1.36740374951033, 0.672415354678118, 1.40774807076906, 1.23085709241133, 1.54303815685287, 1.61726957190717, 1.36121243993766, 0.418990155927757, 1.10541795637635, 0.711390852697397, 0.408418829954349, 0.703152847230079, 0.704742905435769, -0.183833713655054, 1.34003273385856, 0.210120371586764, 0.222845196541972, -0.813324570376352, -0.412441362398562, -1.27382024839739, -0.859352341630058, -1.07915854413014, -0.0302683115662864, 1.76819650181227, 0.0458070914825576, -1.98327828188413, -1.2368391345566, -0.273867626298476, -2.35709998519742, -0.940251320839552, 0.0556024603329693, -0.439361189498476, -0.758666455784151, -0.197579706902746, 0.618366453613662, -0.234761348988804, -1.51306679241434, -1.05135377716523, -0.781881959483692, -0.140531600778617, -1.26428098899066, -0.885158365106794, -0.0578638316665617, 0.906023131624307, 0.463508516204846), Population = c("Algeria (Mzab) – Mozabite", "Bougainville – NAN Melanesian", "Brazil – Karitiana", "Brazil – Surui", "C. African Republic – Biaka Pygmy", "Cambodia – Cambodian", "China – Dai", "China – Daur", "China – Han", "China – Hezhen", "China – Lahu", "China – Miaozu", "China – Mongola", "China – Naxi", "China – Oroqen", "China – She", "China – Tu", "China – Tujia", "China – Uygur", "China – Xibo", "China – Yizu", "Colombia – Piapoco and Curripaco", "D. R. of Congo – Mbuti Pygmy", "France – Basque", "France – French", "Israel (Carmel) – Druze", "Israel (Central) – Palestinian", "Israel (Negev) – Bedouin", "Italy – from Bergamo", "Italy – Sardinian", "Italy – Tuscan", "Japan – Japanese", "Kenya – Bantu", "Mexico – Maya", "Mexico – Pima", "Namibia – San", "New Guinea – Papuan", "Nigeria – Yoruba", "Orkney Islands – Orcadian", "Pakistan – Balochi", "Pakistan – Brahui", "Pakistan – Burusho", "Pakistan – Hazara", "Pakistan – Kalash", "Pakistan – Makrani", "Pakistan – Pathan", "Pakistan – Sindhi", "Population Set 1", "Russia – Russian", "Russia (Caucasus) – Adygei", "Senegal – Mandenka", "Siberia – Yakut", "South Africa – Bantu" )), .Names = c("PGS_Z", "Population"), row.names = c("Algeria (Mzab) – Mozabite", "Bougainville – NAN Melanesian", "Brazil – Karitiana", "Brazil – Surui", "C. African Republic – Biaka Pygmy", "Cambodia – Cambodian", "China – Dai", "China – Daur", "China – Han", "China – Hezhen", "China – Lahu", "China – Miaozu", "China – Mongola", "China – Naxi", "China – Oroqen", "China – She", "China – Tu", "China – Tujia", "China – Uygur", "China – Xibo", "China – Yizu", "Colombia – Piapoco and Curripaco", "D. R. of Congo – Mbuti Pygmy", "France – Basque", "France – French", "Israel (Carmel) – Druze", "Israel (Central) – Palestinian", "Israel (Negev) – Bedouin", "Italy – from Bergamo", "Italy – Sardinian", "Italy – Tuscan", "Japan – Japanese", "Kenya – Bantu", "Mexico – Maya", "Mexico – Pima", "Namibia – San", "New Guinea – Papuan", "Nigeria – Yoruba", "Orkney Islands – Orcadian", "Pakistan – Balochi", "Pakistan – Brahui", "Pakistan – Burusho", "Pakistan – Hazara", "Pakistan – Kalash", "Pakistan – Makrani", "Pakistan – Pathan", "Pakistan – Sindhi", "Population Set 1", "Russia – Russian", "Russia (Caucasus) – Adygei", "Senegal – Mandenka", "Siberia – Yakut", "South Africa – Bantu" ), class = "data.frame")
    df_PGS_cont_20 <- structure(list(PGS_Z = c(0.653893673462433, -0.730733849600686, -0.0670803805416632, 1.69027530476756, -0.0376922725492349, -0.102939430156396, -1.74614356692521, 0.3404205215432), Population = c("AFRICA", "AMERICA", "CENTRAL-SOUTH ASIA", "EAST ASIA", "EUROPE", "MIDDLE EAST", "OCEANIA", "Population Set 1")), .Names = c("PGS_Z", "Population" ), row.names = c("AFRICA", "AMERICA", "CENTRAL-SOUTH ASIA", "EAST ASIA", "EUROPE", "MIDDLE EAST", "OCEANIA", "Population Set 1"), class = "data.frame")

    # Using SNP subset 1e-30 < p < 1e-20
    df_PGS_pop_20_30 <- structure(list(PGS_Z = c(1.50332824545052, 0.00940525814543819, 0.379513465320102, -2.12421853700429, 0.506455789880328, -0.766008854773309, 0.101919149222941, -0.887214805243339, 0.236732431127176, -0.393669269400149, 1.72110498389014, 0.676320728858494, 1.64384563646339, 1.13206019709658, 1.13372778719622, 0.422870807910656, 0.345265036471351, 0.82728687942642, -0.360658796949194, 0.637818416914439, 0.296188301359843, -0.533147388375364, 2.60243314417584, 1.34018121526478, 1.12577531831984, -0.740171397170485, 0.418129553778085, -0.659269503141165, 0.832347387060755, 0.787548309241492, 0.48704585623588, 0.28715071020886, -0.197574598569653, -1.37846103661106, -1.13222828259139, 0.24490396224678, -0.431838742587047, -1.6583835258514, 0.620437780794498, -0.712625593701843, -0.917959936357313, -1.44133190041844, -0.405040128165086, -1.83139345145332, -1.12781816905905, -1.33409498880749, -0.927651317924142, -0.108190410703586, -0.680455708928403, -0.407089803572685, -0.500468056524642, 0.0442497400746086, 1.2929181117484), Population = c("Algeria (Mzab) – Mozabite", "Bougainville – NAN Melanesian", "Brazil – Karitiana", "Brazil – Surui", "C. African Republic – Biaka Pygmy", "Cambodia – Cambodian", "China – Dai", "China – Daur", "China – Han", "China – Hezhen", "China – Lahu", "China – Miaozu", "China – Mongola", "China – Naxi", "China – Oroqen", "China – She", "China – Tu", "China – Tujia", "China – Uygur", "China – Xibo", "China – Yizu", "Colombia – Piapoco and Curripaco", "D. R. of Congo – Mbuti Pygmy", "France – Basque", "France – French", "Israel (Carmel) – Druze", "Israel (Central) – Palestinian", "Israel (Negev) – Bedouin", "Italy – from Bergamo", "Italy – Sardinian", "Italy – Tuscan", "Japan – Japanese", "Kenya – Bantu", "Mexico – Maya", "Mexico – Pima", "Namibia – San", "New Guinea – Papuan", "Nigeria – Yoruba", "Orkney Islands – Orcadian", "Pakistan – Balochi", "Pakistan – Brahui", "Pakistan – Burusho", "Pakistan – Hazara", "Pakistan – Kalash", "Pakistan – Makrani", "Pakistan – Pathan", "Pakistan – Sindhi", "Population Set 1", "Russia – Russian", "Russia (Caucasus) – Adygei", "Senegal – Mandenka", "Siberia – Yakut", "South Africa – Bantu" )), .Names = c("PGS_Z", "Population"), row.names = c("Algeria (Mzab) – Mozabite", "Bougainville – NAN Melanesian", "Brazil – Karitiana", "Brazil – Surui", "C. African Republic – Biaka Pygmy", "Cambodia – Cambodian", "China – Dai", "China – Daur", "China – Han", "China – Hezhen", "China – Lahu", "China – Miaozu", "China – Mongola", "China – Naxi", "China – Oroqen", "China – She", "China – Tu", "China – Tujia", "China – Uygur", "China – Xibo", "China – Yizu", "Colombia – Piapoco and Curripaco", "D. R. of Congo – Mbuti Pygmy", "France – Basque", "France – French", "Israel (Carmel) – Druze", "Israel (Central) – Palestinian", "Israel (Negev) – Bedouin", "Italy – from Bergamo", "Italy – Sardinian", "Italy – Tuscan", "Japan – Japanese", "Kenya – Bantu", "Mexico – Maya", "Mexico – Pima", "Namibia – San", "New Guinea – Papuan", "Nigeria – Yoruba", "Orkney Islands – Orcadian", "Pakistan – Balochi", "Pakistan – Brahui", "Pakistan – Burusho", "Pakistan – Hazara", "Pakistan – Kalash", "Pakistan – Makrani", "Pakistan – Pathan", "Pakistan – Sindhi", "Population Set 1", "Russia – Russian", "Russia (Caucasus) – Adygei", "Senegal – Mandenka", "Siberia – Yakut", "South Africa – Bantu" ), class = "data.frame")
    df_PGS_cont_20_30 <- structure(list(PGS_Z = c(0.438789045972685, -1.35288548955514, -1.55073254279487, 0.927621371934068, 1.25596106648565, 0.316698910098827, -0.147239425870863, 0.11178706372964), Population = c("AFRICA", "AMERICA", "CENTRAL-SOUTH ASIA", "EAST ASIA", "EUROPE", "MIDDLE EAST", "OCEANIA", "Population Set 1")), .Names = c("PGS_Z", "Population" ), row.names = c("AFRICA", "AMERICA", "CENTRAL-SOUTH ASIA", "EAST ASIA", "EUROPE", "MIDDLE EAST", "OCEANIA", "Population Set 1"), class = "data.frame")

    library(corrplot)
    cor(df_PGS_pop_30$PGS_Z, df_PGS_pop_20_30$PGS_Z)
    corrplot(cor(data.frame(df_PGS_pop_20$PGS_Z, df_PGS_pop_30$PGS_Z, df_PGS_pop_20_30$PGS_Z)))

    cor(df_PGS_cont_30$PGS_Z, df_PGS_cont_20_30$PGS_Z)
    corrplot(cor(data.frame(df_PGS_cont_20$PGS_Z, df_PGS_cont_30$PGS_Z, df_PGS_cont_20_30$PGS_Z)))

    Read More
    • Replies: @Factorize
    Thank you for the code.

    I was interested in looking at the different populations chromosome by chromosome. I was able to
    follow along the code and switch in new assignments. I had to give up when I realized that there were only a fairly limited number of genotypes that had been included. While it might make sense to have a very limited number of SNPs in a PGS, I was not sure whether having a small number for each chromosome would be reasonable.

    Your comments have me wondering about the LD question in different populations. Many of the 3,000 SNPs would not be causal. Perhaps a smaller set of SNPs avoids including a lot of this noise.
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  96. res says:
    @Factorize
    res, I was also excited about how academic work could be used as a measure of National achievement.
    Instead of focusing in on teaching to standardized tests such as SAT or PISA, one could have students submit their academic work online (probably best to do this anonymously). This could offer a tremendous insight into the psychometric potential of the entire world. With standardized tests, there is a legitimate question whether having all the students working on the same generic problems offers any marginal utility. Allowing students to research and publish online topics that they were interested in could result in substantial contributions to the knowledge base. At this time most of this youthful energy is not captured.

    In addition, having an online repository of academic output could be a great way to document the psychometric changes that occur as technology has advanced.

    That is an interesting idea. It also offers great potential as a talent search (e.g. though anonymous, allow email contact through an intermediary).

    Read More
    • Replies: @Factorize
    res, thank you for replying to this comment because I think it is quite a good idea. It is not clear to me why this hasn't been done. One concern that I would have is that it could drive unhealthy competitive pressures into the school system that might ultimately diminish not contribute to higher performance. Yet, some of standardized testing has already been accused of this. It could also be a problem if difficult subjects became highly politicized when students were still early in their educations. Psychometrics would clearly be one such topic.

    res, I am preparing to start a course on learning. Any inputs you (or others on the blog) might offer on operant or classical conditioning would be appreciated.

    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  97. Factorize says:
    @res
    That is an interesting idea. It also offers great potential as a talent search (e.g. though anonymous, allow email contact through an intermediary).

    res, thank you for replying to this comment because I think it is quite a good idea. It is not clear to me why this hasn’t been done. One concern that I would have is that it could drive unhealthy competitive pressures into the school system that might ultimately diminish not contribute to higher performance. Yet, some of standardized testing has already been accused of this. It could also be a problem if difficult subjects became highly politicized when students were still early in their educations. Psychometrics would clearly be one such topic.

    res, I am preparing to start a course on learning. Any inputs you (or others on the blog) might offer on operant or classical conditioning would be appreciated.

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

    With the current article, I had quite a bit of fun forking the R code, even though I have little understanding of R. All I had to do was take the code that was given and remix it.
     
    I did the same thing. One thing I did was calculate -log10(p) (a standard measure of significance in the GWAS literature) for the SNPs which makes it much easier to look at things like histograms of SNP significance by p-value. Another thing was to look at SNPs in the exclusive intervals with p < 10^-30 and 10^-30 < p < 10^-20. I was surprised to find there was almost 0 (|| < 0.06 for both populations and continents) correlation between the PGS computed from those subsets. I think this calls into question the "small number of SNPs show selection pressure so conclusions are applicable to the other SNPs" idea. Though perhaps the more limited version in Piffer's work is still applicable. Namely:
    1. Very small number of SNPs (e.g. 9) which would most clearly show selection pressure.
    2. Limiting to SNPs with positive effect for the derived allele.

    The subtlety of 2. is something I only really noticed recently in Piffer's early work. I initially worried about it as cherry picking, but I am coming to think of it as sensible. The interesting question there is what is happening with the SNPs that have a negative effect for the derived allele yet have managed to achieve a somewhat high frequency. IIRC those are the SNPs where the lower IQ variant is non-intuitively more frequent in higher IQ populations.

    It is important to note that the low correlation is seen only with exclusive subsets. PGS from p < 10^-30 and p < 10^-20 correlate at 0.77. This is what I was alluding to in my A + B statement in comment 85.

    To draw more firm conclusions it would be good to look at the SNP characteristics more closely by subset. Are the derived alleles typically + or -, how much (and with what pattern) do the frequencies vary between populations, etc.?

    If you want to learn R, how to use it effectively, and see the breadth of tools and capabilities (e.g. RStudio, R Markdown, RPubs, Shiny interactive web apps, GitHub) available I highly recommend this series of courses: https://www.coursera.org/specializations/jhu-data-science

    It looks like they are no longer free, but if you are a tuition paying student I think you will see they offer a great deal of value for the money ($49 per class).

    I think course 5, Reproducible Research, is especially relevant to the conversation we are having now: https://www.coursera.org/learn/reproducible-research

    P.S. Some fairly simple code you might find helpful.

    gwas$log10p 7)], breaks=7:ceiling(max(gwas$log10p))) # Roughly significant SNPs
     

    # Use ranges instead
    hpthresh <- 10^-20
    lpthresh <- 10^-30
    gwas=gwas[which(gwas$Pvallpthresh),] # select SNPs within range
    subset.note <- paste(lpthresh, "< p <", hpthresh)

    hist(gwas$log10p, breaks=max(gwas$log10p), main=paste("SNP Frequency Distribution for", subset.note))

    cat("Using SNP subset", subset.note)
     
    I then output the PGS results for both populations and continents so I could load them as below. There is an annoying issue where R Markdown uses UTF-8 style double quotes which I had to convert to " below.

    ```{r, results='asis'}
    cat("Using SNP subset", subset.note)
    # Formatted data for cut and paste from knitted HTML
    dump("df_PGS", "")
    ```

    If you want to check my work, here is how I generated my results:

    # Using SNP subset p < 5e-08
    df_PGS <- structure(list(PGS_Z = c(0.5200636075001, -1.11902371997781, -1.02208247762685, -2.07651476702638, -0.302255484493313, -1.42211430460949, 1.40980466840254, 0.559582471229181, 1.09486218318298, 0.504133511124049, -0.466988474444203, 0.967393318105655, 1.69173244649728, 1.12514986737479, 1.67521351362267, 0.326979905514463, -0.310754319990793, 1.32151715857769, 0.0685786322416409, 1.14807956305697, 0.788887780341753, -0.372752615000197, 0.860232668360736, 1.3287069623234, 0.92922372898383, -0.581182512307871, -0.224925800495491, -0.764091225900405, 0.185373269728186, -0.7909882596934, 0.369592154021238, 0.248155541120252, -0.530569528552552, -1.28424949508351, -0.946197739930438, -1.72471514128299, -3.18629540939235, -1.19153830835272, 1.02634720313318, 0.338840917692266, -0.222976299902283, 0.568804159526946, 0.68520661899737, 0.40549941129595, -0.32654166812477, -0.594762957433412, 0.0515190783651431, -0.0106906814603098, 0.308741160042542, -0.246360815781283, -0.787651899184517, 0.755315367983932, -0.757312962299404), Population = c("Algeria (Mzab) - Mozabite", "Bougainville - NAN Melanesian", "Brazil - Karitiana", "Brazil - Surui", "C. African Republic - Biaka Pygmy", "Cambodia - Cambodian", "China - Dai", "China - Daur", "China - Han", "China - Hezhen", "China - Lahu", "China - Miaozu", "China - Mongola", "China - Naxi", "China - Oroqen", "China - She", "China - Tu", "China - Tujia", "China - Uygur", "China - Xibo", "China - Yizu", "Colombia - Piapoco and Curripaco", "D. R. of Congo - Mbuti Pygmy", "France - Basque", "France - French", "Israel (Carmel) - Druze", "Israel (Central) - Palestinian", "Israel (Negev) - Bedouin", "Italy - from Bergamo", "Italy - Sardinian", "Italy - Tuscan", "Japan - Japanese", "Kenya - Bantu", "Mexico - Maya", "Mexico - Pima", "Namibia - San", "New Guinea - Papuan", "Nigeria - Yoruba", "Orkney Islands - Orcadian", "Pakistan - Balochi", "Pakistan - Brahui", "Pakistan - Burusho", "Pakistan - Hazara", "Pakistan - Kalash", "Pakistan - Makrani", "Pakistan - Pathan", "Pakistan - Sindhi", "Population Set 1", "Russia - Russian", "Russia (Caucasus) - Adygei", "Senegal - Mandenka", "Siberia - Yakut", "South Africa - Bantu" )), .Names = c("PGS_Z", "Population"), row.names = c("Algeria (Mzab) - Mozabite", "Bougainville - NAN Melanesian", "Brazil - Karitiana", "Brazil - Surui", "C. African Republic - Biaka Pygmy", "Cambodia - Cambodian", "China - Dai", "China - Daur", "China - Han", "China - Hezhen", "China - Lahu", "China - Miaozu", "China - Mongola", "China - Naxi", "China - Oroqen", "China - She", "China - Tu", "China - Tujia", "China - Uygur", "China - Xibo", "China - Yizu", "Colombia - Piapoco and Curripaco", "D. R. of Congo - Mbuti Pygmy", "France - Basque", "France - French", "Israel (Carmel) - Druze", "Israel (Central) - Palestinian", "Israel (Negev) - Bedouin", "Italy - from Bergamo", "Italy - Sardinian", "Italy - Tuscan", "Japan - Japanese", "Kenya - Bantu", "Mexico - Maya", "Mexico - Pima", "Namibia - San", "New Guinea - Papuan", "Nigeria - Yoruba", "Orkney Islands - Orcadian", "Pakistan - Balochi", "Pakistan - Brahui", "Pakistan - Burusho", "Pakistan - Hazara", "Pakistan - Kalash", "Pakistan - Makrani", "Pakistan - Pathan", "Pakistan - Sindhi", "Population Set 1", "Russia - Russian", "Russia (Caucasus) - Adygei", "Senegal - Mandenka", "Siberia - Yakut", "South Africa - Bantu" ), class = "data.frame")
    df_PGS <- structure(list(PGS_Z = c(-0.171877416500972, -0.76325883175675, 0.527052463582658, 1.13366534856215, 0.802379817204429, 0.0721060622819543, -2.00331924902356, 0.403251805650091), Population = c("AFRICA", "AMERICA", "CENTRAL-SOUTH ASIA", "EAST ASIA", "EUROPE", "MIDDLE EAST", "OCEANIA", "Population Set 1")), .Names = c("PGS_Z", "Population" ), row.names = c("AFRICA", "AMERICA", "CENTRAL-SOUTH ASIA", "EAST ASIA", "EUROPE", "MIDDLE EAST", "OCEANIA", "Population Set 1"), class = "data.frame")

    # Using SNP subset p < 1e-30
    df_PGS_pop_30 <- structure(list(PGS_Z = c(-0.22774097619992, -1.9715591510291, -0.671060706737492, 0.640248446937808, 0.498481462070653, -0.165132021585292, 1.20605525284874, 1.59005582149218, 1.5507521144634, 1.18022912859246, 0.38682595275503, 1.01660713020854, 0.622694079610597, 1.13644957931209, 0.808396343591038, 0.188319081351515, 1.12763415841273, 0.230240996196302, 0.816383059308068, 0.374866735853536, 0.656620936068732, 0.201996888194904, -0.421196361148007, -0.829255798533044, -0.637466166619254, -0.431621917284199, -0.86857201956625, -1.08538270811792, -1.77791883690442, -2.02167495788651, -0.437408949111198, 2.02081705336049, 0.22021588820191, -1.40166582744889, -0.650943749391742, -0.549937657847137, -2.65369336124334, 0.158268785377203, -0.43707246143649, 0.0229409510124096, -0.216314299709225, 0.928077501081921, 1.12058182169323, 1.20002072182299, -1.0069769085762, -0.249006441063474, -0.237998288498758, -0.0907098467875349, -1.0558649525544, -0.795996355644576, 0.335956524741878, 1.11970862227292, -0.467274315908915), Population = c("Algeria (Mzab) - Mozabite", "Bougainville - NAN Melanesian", "Brazil - Karitiana", "Brazil - Surui", "C. African Republic - Biaka Pygmy", "Cambodia - Cambodian", "China - Dai", "China - Daur", "China - Han", "China - Hezhen", "China - Lahu", "China - Miaozu", "China - Mongola", "China - Naxi", "China - Oroqen", "China - She", "China - Tu", "China - Tujia", "China - Uygur", "China - Xibo", "China - Yizu", "Colombia - Piapoco and Curripaco", "D. R. of Congo - Mbuti Pygmy", "France - Basque", "France - French", "Israel (Carmel) - Druze", "Israel (Central) - Palestinian", "Israel (Negev) - Bedouin", "Italy - from Bergamo", "Italy - Sardinian", "Italy - Tuscan", "Japan - Japanese", "Kenya - Bantu", "Mexico - Maya", "Mexico - Pima", "Namibia - San", "New Guinea - Papuan", "Nigeria - Yoruba", "Orkney Islands - Orcadian", "Pakistan - Balochi", "Pakistan - Brahui", "Pakistan - Burusho", "Pakistan - Hazara", "Pakistan - Kalash", "Pakistan - Makrani", "Pakistan - Pathan", "Pakistan - Sindhi", "Population Set 1", "Russia - Russian", "Russia (Caucasus) - Adygei", "Senegal - Mandenka", "Siberia - Yakut", "South Africa - Bantu" )), .Names = c("PGS_Z", "Population"), row.names = c("Algeria (Mzab) - Mozabite", "Bougainville - NAN Melanesian", "Brazil - Karitiana", "Brazil - Surui", "C. African Republic - Biaka Pygmy", "Cambodia - Cambodian", "China - Dai", "China - Daur", "China - Han", "China - Hezhen", "China - Lahu", "China - Miaozu", "China - Mongola", "China - Naxi", "China - Oroqen", "China - She", "China - Tu", "China - Tujia", "China - Uygur", "China - Xibo", "China - Yizu", "Colombia - Piapoco and Curripaco", "D. R. of Congo - Mbuti Pygmy", "France - Basque", "France - French", "Israel (Carmel) - Druze", "Israel (Central) - Palestinian", "Israel (Negev) - Bedouin", "Italy - from Bergamo", "Italy - Sardinian", "Italy - Tuscan", "Japan - Japanese", "Kenya - Bantu", "Mexico - Maya", "Mexico - Pima", "Namibia - San", "New Guinea - Papuan", "Nigeria - Yoruba", "Orkney Islands - Orcadian", "Pakistan - Balochi", "Pakistan - Brahui", "Pakistan - Burusho", "Pakistan - Hazara", "Pakistan - Kalash", "Pakistan - Makrani", "Pakistan - Pathan", "Pakistan - Sindhi", "Population Set 1", "Russia - Russian", "Russia (Caucasus) - Adygei", "Senegal - Mandenka", "Siberia - Yakut", "South Africa - Bantu" ), class = "data.frame")
    df_PGS_cont_30 <- structure(list(PGS_Z = c(0.534009086642049, -0.207818862875531, 0.624192430205281, 1.47355288841214, -0.608538124342132, -0.258024397998142, -1.88798149112997, 0.330608471086306), Population = c("AFRICA", "AMERICA", "CENTRAL-SOUTH ASIA", "EAST ASIA", "EUROPE", "MIDDLE EAST", "OCEANIA", "Population Set 1")), .Names = c("PGS_Z", "Population" ), row.names = c("AFRICA", "AMERICA", "CENTRAL-SOUTH ASIA", "EAST ASIA", "EUROPE", "MIDDLE EAST", "OCEANIA", "Population Set 1"), class = "data.frame")

    # Using SNP subset p < 1e-20
    df_PGS_pop_20 <- structure(list(PGS_Z = c(0.786289242132707, -1.53926733341709, -0.282415142753057, -0.861436091656981, 0.715737084448114, -0.621032150920052, 1.01071195358365, 0.676893455546647, 1.36740374951033, 0.672415354678118, 1.40774807076906, 1.23085709241133, 1.54303815685287, 1.61726957190717, 1.36121243993766, 0.418990155927757, 1.10541795637635, 0.711390852697397, 0.408418829954349, 0.703152847230079, 0.704742905435769, -0.183833713655054, 1.34003273385856, 0.210120371586764, 0.222845196541972, -0.813324570376352, -0.412441362398562, -1.27382024839739, -0.859352341630058, -1.07915854413014, -0.0302683115662864, 1.76819650181227, 0.0458070914825576, -1.98327828188413, -1.2368391345566, -0.273867626298476, -2.35709998519742, -0.940251320839552, 0.0556024603329693, -0.439361189498476, -0.758666455784151, -0.197579706902746, 0.618366453613662, -0.234761348988804, -1.51306679241434, -1.05135377716523, -0.781881959483692, -0.140531600778617, -1.26428098899066, -0.885158365106794, -0.0578638316665617, 0.906023131624307, 0.463508516204846), Population = c("Algeria (Mzab) - Mozabite", "Bougainville - NAN Melanesian", "Brazil - Karitiana", "Brazil - Surui", "C. African Republic - Biaka Pygmy", "Cambodia - Cambodian", "China - Dai", "China - Daur", "China - Han", "China - Hezhen", "China - Lahu", "China - Miaozu", "China - Mongola", "China - Naxi", "China - Oroqen", "China - She", "China - Tu", "China - Tujia", "China - Uygur", "China - Xibo", "China - Yizu", "Colombia - Piapoco and Curripaco", "D. R. of Congo - Mbuti Pygmy", "France - Basque", "France - French", "Israel (Carmel) - Druze", "Israel (Central) - Palestinian", "Israel (Negev) - Bedouin", "Italy - from Bergamo", "Italy - Sardinian", "Italy - Tuscan", "Japan - Japanese", "Kenya - Bantu", "Mexico - Maya", "Mexico - Pima", "Namibia - San", "New Guinea - Papuan", "Nigeria - Yoruba", "Orkney Islands - Orcadian", "Pakistan - Balochi", "Pakistan - Brahui", "Pakistan - Burusho", "Pakistan - Hazara", "Pakistan - Kalash", "Pakistan - Makrani", "Pakistan - Pathan", "Pakistan - Sindhi", "Population Set 1", "Russia - Russian", "Russia (Caucasus) - Adygei", "Senegal - Mandenka", "Siberia - Yakut", "South Africa - Bantu" )), .Names = c("PGS_Z", "Population"), row.names = c("Algeria (Mzab) - Mozabite", "Bougainville - NAN Melanesian", "Brazil - Karitiana", "Brazil - Surui", "C. African Republic - Biaka Pygmy", "Cambodia - Cambodian", "China - Dai", "China - Daur", "China - Han", "China - Hezhen", "China - Lahu", "China - Miaozu", "China - Mongola", "China - Naxi", "China - Oroqen", "China - She", "China - Tu", "China - Tujia", "China - Uygur", "China - Xibo", "China - Yizu", "Colombia - Piapoco and Curripaco", "D. R. of Congo - Mbuti Pygmy", "France - Basque", "France - French", "Israel (Carmel) - Druze", "Israel (Central) - Palestinian", "Israel (Negev) - Bedouin", "Italy - from Bergamo", "Italy - Sardinian", "Italy - Tuscan", "Japan - Japanese", "Kenya - Bantu", "Mexico - Maya", "Mexico - Pima", "Namibia - San", "New Guinea - Papuan", "Nigeria - Yoruba", "Orkney Islands - Orcadian", "Pakistan - Balochi", "Pakistan - Brahui", "Pakistan - Burusho", "Pakistan - Hazara", "Pakistan - Kalash", "Pakistan - Makrani", "Pakistan - Pathan", "Pakistan - Sindhi", "Population Set 1", "Russia - Russian", "Russia (Caucasus) - Adygei", "Senegal - Mandenka", "Siberia - Yakut", "South Africa - Bantu" ), class = "data.frame")
    df_PGS_cont_20 <- structure(list(PGS_Z = c(0.653893673462433, -0.730733849600686, -0.0670803805416632, 1.69027530476756, -0.0376922725492349, -0.102939430156396, -1.74614356692521, 0.3404205215432), Population = c("AFRICA", "AMERICA", "CENTRAL-SOUTH ASIA", "EAST ASIA", "EUROPE", "MIDDLE EAST", "OCEANIA", "Population Set 1")), .Names = c("PGS_Z", "Population" ), row.names = c("AFRICA", "AMERICA", "CENTRAL-SOUTH ASIA", "EAST ASIA", "EUROPE", "MIDDLE EAST", "OCEANIA", "Population Set 1"), class = "data.frame")

    # Using SNP subset 1e-30 < p < 1e-20
    df_PGS_pop_20_30 <- structure(list(PGS_Z = c(1.50332824545052, 0.00940525814543819, 0.379513465320102, -2.12421853700429, 0.506455789880328, -0.766008854773309, 0.101919149222941, -0.887214805243339, 0.236732431127176, -0.393669269400149, 1.72110498389014, 0.676320728858494, 1.64384563646339, 1.13206019709658, 1.13372778719622, 0.422870807910656, 0.345265036471351, 0.82728687942642, -0.360658796949194, 0.637818416914439, 0.296188301359843, -0.533147388375364, 2.60243314417584, 1.34018121526478, 1.12577531831984, -0.740171397170485, 0.418129553778085, -0.659269503141165, 0.832347387060755, 0.787548309241492, 0.48704585623588, 0.28715071020886, -0.197574598569653, -1.37846103661106, -1.13222828259139, 0.24490396224678, -0.431838742587047, -1.6583835258514, 0.620437780794498, -0.712625593701843, -0.917959936357313, -1.44133190041844, -0.405040128165086, -1.83139345145332, -1.12781816905905, -1.33409498880749, -0.927651317924142, -0.108190410703586, -0.680455708928403, -0.407089803572685, -0.500468056524642, 0.0442497400746086, 1.2929181117484), Population = c("Algeria (Mzab) - Mozabite", "Bougainville - NAN Melanesian", "Brazil - Karitiana", "Brazil - Surui", "C. African Republic - Biaka Pygmy", "Cambodia - Cambodian", "China - Dai", "China - Daur", "China - Han", "China - Hezhen", "China - Lahu", "China - Miaozu", "China - Mongola", "China - Naxi", "China - Oroqen", "China - She", "China - Tu", "China - Tujia", "China - Uygur", "China - Xibo", "China - Yizu", "Colombia - Piapoco and Curripaco", "D. R. of Congo - Mbuti Pygmy", "France - Basque", "France - French", "Israel (Carmel) - Druze", "Israel (Central) - Palestinian", "Israel (Negev) - Bedouin", "Italy - from Bergamo", "Italy - Sardinian", "Italy - Tuscan", "Japan - Japanese", "Kenya - Bantu", "Mexico - Maya", "Mexico - Pima", "Namibia - San", "New Guinea - Papuan", "Nigeria - Yoruba", "Orkney Islands - Orcadian", "Pakistan - Balochi", "Pakistan - Brahui", "Pakistan - Burusho", "Pakistan - Hazara", "Pakistan - Kalash", "Pakistan - Makrani", "Pakistan - Pathan", "Pakistan - Sindhi", "Population Set 1", "Russia - Russian", "Russia (Caucasus) - Adygei", "Senegal - Mandenka", "Siberia - Yakut", "South Africa - Bantu" )), .Names = c("PGS_Z", "Population"), row.names = c("Algeria (Mzab) - Mozabite", "Bougainville - NAN Melanesian", "Brazil - Karitiana", "Brazil - Surui", "C. African Republic - Biaka Pygmy", "Cambodia - Cambodian", "China - Dai", "China - Daur", "China - Han", "China - Hezhen", "China - Lahu", "China - Miaozu", "China - Mongola", "China - Naxi", "China - Oroqen", "China - She", "China - Tu", "China - Tujia", "China - Uygur", "China - Xibo", "China - Yizu", "Colombia - Piapoco and Curripaco", "D. R. of Congo - Mbuti Pygmy", "France - Basque", "France - French", "Israel (Carmel) - Druze", "Israel (Central) - Palestinian", "Israel (Negev) - Bedouin", "Italy - from Bergamo", "Italy - Sardinian", "Italy - Tuscan", "Japan - Japanese", "Kenya - Bantu", "Mexico - Maya", "Mexico - Pima", "Namibia - San", "New Guinea - Papuan", "Nigeria - Yoruba", "Orkney Islands - Orcadian", "Pakistan - Balochi", "Pakistan - Brahui", "Pakistan - Burusho", "Pakistan - Hazara", "Pakistan - Kalash", "Pakistan - Makrani", "Pakistan - Pathan", "Pakistan - Sindhi", "Population Set 1", "Russia - Russian", "Russia (Caucasus) - Adygei", "Senegal - Mandenka", "Siberia - Yakut", "South Africa - Bantu" ), class = "data.frame")
    df_PGS_cont_20_30 <- structure(list(PGS_Z = c(0.438789045972685, -1.35288548955514, -1.55073254279487, 0.927621371934068, 1.25596106648565, 0.316698910098827, -0.147239425870863, 0.11178706372964), Population = c("AFRICA", "AMERICA", "CENTRAL-SOUTH ASIA", "EAST ASIA", "EUROPE", "MIDDLE EAST", "OCEANIA", "Population Set 1")), .Names = c("PGS_Z", "Population" ), row.names = c("AFRICA", "AMERICA", "CENTRAL-SOUTH ASIA", "EAST ASIA", "EUROPE", "MIDDLE EAST", "OCEANIA", "Population Set 1"), class = "data.frame")

    library(corrplot)
    cor(df_PGS_pop_30$PGS_Z, df_PGS_pop_20_30$PGS_Z)
    corrplot(cor(data.frame(df_PGS_pop_20$PGS_Z, df_PGS_pop_30$PGS_Z, df_PGS_pop_20_30$PGS_Z)))

    cor(df_PGS_cont_30$PGS_Z, df_PGS_cont_20_30$PGS_Z)
    corrplot(cor(data.frame(df_PGS_cont_20$PGS_Z, df_PGS_cont_30$PGS_Z, df_PGS_cont_20_30$PGS_Z)))

     

    Thank you for the code.

    I was interested in looking at the different populations chromosome by chromosome. I was able to
    follow along the code and switch in new assignments. I had to give up when I realized that there were only a fairly limited number of genotypes that had been included. While it might make sense to have a very limited number of SNPs in a PGS, I was not sure whether having a small number for each chromosome would be reasonable.

    Your comments have me wondering about the LD question in different populations. Many of the 3,000 SNPs would not be causal. Perhaps a smaller set of SNPs avoids including a lot of this noise.

    Read More
    • Replies: @res

    Your comments have me wondering about the LD question in different populations. Many of the 3,000 SNPs would not be causal. Perhaps a smaller set of SNPs avoids including a lot of this noise.
     
    I don't know if there is any good reason to worry more about LD in the more vs. less significant SNPs.

    Did you see Piffer's LD decay work last year?
    http://rpubs.com/Daxide/283453
    ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
  99. res says:
    @Factorize
    Thank you for the code.

    I was interested in looking at the different populations chromosome by chromosome. I was able to
    follow along the code and switch in new assignments. I had to give up when I realized that there were only a fairly limited number of genotypes that had been included. While it might make sense to have a very limited number of SNPs in a PGS, I was not sure whether having a small number for each chromosome would be reasonable.

    Your comments have me wondering about the LD question in different populations. Many of the 3,000 SNPs would not be causal. Perhaps a smaller set of SNPs avoids including a lot of this noise.

    Your comments have me wondering about the LD question in different populations. Many of the 3,000 SNPs would not be causal. Perhaps a smaller set of SNPs avoids including a lot of this noise.

    I don’t know if there is any good reason to worry more about LD in the more vs. less significant SNPs.

    Did you see Piffer’s LD decay work last year?

    http://rpubs.com/Daxide/283453

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

    res, thank you very much for the reference!

    Read More
    ReplyAgree/Disagree/Etc. More... This Commenter Display All Comments
  101. Factorize says:

    res, could you provide your take on the results from Head Start? Program appeared to increase IQ by 9 points but this faded out through time. Was this simply because there were no natural reinforcers to sustain the behavior? Have there ever been attempts to create such ongoing reinforcement? It seems quite surprising that possibly 10 IQ points are more could just be allowed to extinct without anyone stepping forward to claim the latest cognitive wealth. If instead of this being poor disadvantaged children, this were children in the top 1% of the distribution there is little chance that such cognitive improvement would be allowed to melt away.

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

Leave a Reply - Comments on articles more than two weeks old will be judged much more strictly on quality and tone


 Remember My InformationWhy?
 Email Replies to my Comment
Submitted comments become the property of The Unz Review and may be republished elsewhere at the sole discretion of the latter
Subscribe to This Comment Thread via RSS Subscribe to All James Thompson Comments via RSS