The Unz Review: An Alternative Media Selection
A Collection of Interesting, Important, and Controversial Perspectives Largely Excluded from the American Mainstream Media
 TeasersiSteve Blog
Plomin: "Polygenic Scores for Educational Attainment and Intelligence Are the Most Powerful Predictors in the Behavioural Sciences"
🔊 Listen RSS
Email This Page to Someone

 Remember My Information



=>

Bookmark Toggle AllToCAdd to LibraryRemove from Library • BShow CommentNext New CommentNext New ReplyRead More
ReplyAgree/Disagree/Etc. More... This Commenter This Thread Hide Thread Display All Comments
AgreeDisagreeThanksLOLTroll
These buttons register your public Agreement, Disagreement, Thanks, LOL, or Troll with the selected comment. They are ONLY available to recent, frequent commenters who have saved their Name+Email using the 'Remember My Information' checkbox, and may also ONLY be used three times during any eight hour period.
Ignore Commenter Follow Commenter
Search Text Case Sensitive  Exact Words  Include Comments
List of Bookmarks

From a recent preprint scientific paper:

Genomic prediction of cognitive traits in childhood and adolescence

Andrea Allegrini, Saskia Selzam, Kaili Rimfeld, Sophie von Stumm, Jean-Baptiste Pingault, Robert Plomin

This article is a preprint and has not been peer-reviewed

Abstract
Recent advances in genomics are producing powerful DNA predictors of complex traits, especially cognitive abilities. Here, we leveraged summary statistics from the most recent genome-wide association studies of intelligence and educational attainment to build prediction models of general cognitive ability and educational achievement. To this end, we compared the performances of multi-trait genomic and polygenic scoring methods. In a representative UK sample of 7,026 children at age 12 and 16, we show that we can now predict up to 11 percent of the variance in intelligence and 16 percent in educational achievement. We also show that predictive power increases from age 12 to age 16 and that genomic predictions do not differ for girls and boys. Multivariate genomic methods were effective in boosting predictive power and, even though prediction accuracy varied across polygenic scores approaches, results were similar using different multivariate and polygenic score methods. Polygenic scores for educational attainment and intelligence are the most powerful predictors in the behavioural sciences and exceed predictions that can be made from parental phenotypes such as educational attainment and occupational status.

 
Hide 29 CommentsLeave a Comment
Commenters to Ignore...to FollowEndorsed Only
Trim Comments?
  1. the Most Powerful Predictors in the Behavioural Sciences

    Okay, I just know this is gonna be racist.

    • LOL: bomag
  2. predictive power increases from age 12 to age 16

    This is the point I tried to make in my criticism of Chanda’s article. Cognitive ability in children is malleable and is thus a poor predictor of adult intelligence. We can do a lot to increase IQ in children, but this increase tends to wash out during adolescence and is largely gone by one’s early twenties.

    It would be interesting to see whether this “adolescence wash-out” differs between human populations, both in its timing and in its magnitude.

    • Replies: @Steve Sailer
    If you can boost IQ temporarily in kids, can you teach them more stuff that they will retain as adults?

    For example, literacy is a big deal. If you get your IQ boosted high enough that you learn how to read, you probably won't forget as an adult, and it's a lot better to go through life as a pretty dumb person who can read than a pretty dumb person who is illiterate.

    , @Joe Schmoe
    Are we really talking IQ boost?

    If it doesn't translate to being permanent, it is kind of like discounts at the store bringing future sales to the present but not increasing the total sales over time.

    What am I missing?

  3. @Peter Frost
    predictive power increases from age 12 to age 16

    This is the point I tried to make in my criticism of Chanda's article. Cognitive ability in children is malleable and is thus a poor predictor of adult intelligence. We can do a lot to increase IQ in children, but this increase tends to wash out during adolescence and is largely gone by one's early twenties.

    It would be interesting to see whether this "adolescence wash-out" differs between human populations, both in its timing and in its magnitude.

    If you can boost IQ temporarily in kids, can you teach them more stuff that they will retain as adults?

    For example, literacy is a big deal. If you get your IQ boosted high enough that you learn how to read, you probably won’t forget as an adult, and it’s a lot better to go through life as a pretty dumb person who can read than a pretty dumb person who is illiterate.

    • Replies: @Jon
    My understanding of it was that you could get someone to learn the things they were capable of learning sooner through better/more focused education, but you couldn't do much to push the upper boundaries of what they were actually capable of learning. That's why the tests for kids under 12 aren't very predictable - almost everyone is capable of learning elementary school stuff, some just get pushed into it sooner and look like baby geniuses - but becomes more predictable after 16 - now we are sorting kids based on whether or not they can understand intro level calc and physics. So literacy would still be limited to those who were capable in the first place, and it wouldn't matter if we taught this early because they should still be able to pick it up anytime.
    , @anon
    I taught my daughter her alphabet with those magnetic letters which was sort of reading as opposed to reciting alphabet at as soon as she could speak which was quite early. It was quite easy,, her mom and I are +2 sds what was interesting is after a few months of no lessons she forget them all again and i had to start over, again not too hard but she again forgot them without continuing training. cant say what would have happened if i was more diligent and kept increasing lessons without stopping, but I suspected I could have gone much farther but i also suspected before they are developed fully to whatever level you are training youre sort of writing on sand
    , @Desiderius
    Can read + has read = the real difference maker.

    If all one reads is Twitter and Potter that Can Read is wasted.

    Terrible thing!
  4. We’ve had EPDs (Expected Progeny Difference) for choosing which bull’s semen to use for breeding for decades now. If humans are just another mammal……..

    • Replies: @bomag

    EPDs...
     
    A powerful tool. At cocktail parties, I used to excitedly explain how great they are, and was looking forward to the day when we could select dating partners for our kids based on the same techniques.

    I noticed people got uncomfortable, so I've dropped the subject. But thinking back to some pop culture references, it was more of a thing back in the day. On the Michael J Fox show Family Ties, he tries to get Mallory to date one of his clean cut friends by saying: "look at this guy; he is genetically incapable of making less than $50,000 a year." In the Ian Anderson song Heavy Horses he croons: "let me find you a filly for your fine stallion seed; to keep... the old line... going."

    Certainly don't hear such musings in today's pop world.
  5. “we can now predict up to 11 percent of the variance in intelligence and 16 percent in educational achievement.” Which sounds quite small. On the other hand it’s much bigger than a few years ago.

    “the most powerful predictors in the behavioural sciences”: setting the bar low, eh?

    I do wish people wouldn’t say absurd corporate things such as “we leveraged ” when all they mean is “we used”.

  6. @Steve Sailer
    If you can boost IQ temporarily in kids, can you teach them more stuff that they will retain as adults?

    For example, literacy is a big deal. If you get your IQ boosted high enough that you learn how to read, you probably won't forget as an adult, and it's a lot better to go through life as a pretty dumb person who can read than a pretty dumb person who is illiterate.

    My understanding of it was that you could get someone to learn the things they were capable of learning sooner through better/more focused education, but you couldn’t do much to push the upper boundaries of what they were actually capable of learning. That’s why the tests for kids under 12 aren’t very predictable – almost everyone is capable of learning elementary school stuff, some just get pushed into it sooner and look like baby geniuses – but becomes more predictable after 16 – now we are sorting kids based on whether or not they can understand intro level calc and physics. So literacy would still be limited to those who were capable in the first place, and it wouldn’t matter if we taught this early because they should still be able to pick it up anytime.

    • Replies: @Steve Sailer
    But literacy percentages seem like the kind of thing that differs radically across time and place. So the quantity and quality of schooling seems to matter whether a kid learns to read or not.
  7. @Jon
    My understanding of it was that you could get someone to learn the things they were capable of learning sooner through better/more focused education, but you couldn't do much to push the upper boundaries of what they were actually capable of learning. That's why the tests for kids under 12 aren't very predictable - almost everyone is capable of learning elementary school stuff, some just get pushed into it sooner and look like baby geniuses - but becomes more predictable after 16 - now we are sorting kids based on whether or not they can understand intro level calc and physics. So literacy would still be limited to those who were capable in the first place, and it wouldn't matter if we taught this early because they should still be able to pick it up anytime.

    But literacy percentages seem like the kind of thing that differs radically across time and place. So the quantity and quality of schooling seems to matter whether a kid learns to read or not.

    • Replies: @megabar
    > If you can boost IQ temporarily in kids, can you teach them more stuff that they will retain as adults?

    Yes, it's interesting that the following appear to be true:
    * IQ in adulthood is primarily genetic, and not subject to improvements via childhood training.
    * Brain plasticity is higher in children, and allows easier or more complete acquisition of some types of skills.

    These two statements are in some ways contradictory. You would assume that the higher plasticity of brains in childhood _would_ allow some learning of things that would be useful in IQ tests. But perhaps brain plasticity is only useful for the kinds of skills that are irrelevant in g-loaded tests.

    Perhaps learning in the critical window is a source of some of the Flynn-effect, or differences between African and American black IQ results.
    , @Diversity Heretic
    Wouldn't a confounding factor be the need for literacy? A medieval peasant with a genius level IQ would never have a need to learn to read and write; the rudiments of farming and even possibly how to improve farming efficiency could be mastered without that fancy "book learning." It's hard for people who read and write a lot to imagine anyone with any significant cognitive ability living as an illiterate, but in some cultures it might be possible. Illiterate people sometimes have prodigious memory capacity--if you can write things down you just don't need that.
    , @bomag

    literacy percentages seem like the kind of thing that differs radically across time and place. So the quantity and quality of schooling...
     
    Isn't is axiomatic that anyone who can walk and talk, sans a disability, can learn to read and write if they deem it something important?

    My old school neighbor with a keen business sense runs a moderately successful enterprise, yet is essentially illiterate. He grew up where schooling was shunned: no books, no practice of writing to record events, etc. (Maybe it helped him hone his memory). Today's world has made letters a much more salient thing, and his kids and grandkids are as literate as one would expect, and I'm sure he could have learned to read on his own if it was an important feature of his upbringing.
  8. Anon[147] • Disclaimer says:

    Polygenic scores for educational attainment and intelligence are the most powerful predictors in the behavioural sciences and exceed predictions that can be made from parental phenotypes such as educational attainment and occupational status.

    Isn’t this circular reasoning? Polygenic scores were derived from genetic data where the smart people were assumed from educational attainment. The research releases signed by the DBA donors only had educational attainment and/or highest math class completed, not IQ data. The IQ was derived from educational attainment, which is a very noisy cognitive measure, but as n approaches infinity the noise approaches zero. So to circle back and say that polygenic scores (based on educational attainment) are better than educational attainment doesn’t make sense to me.

    If you can boost IQ temporarily in kids, can you teach them more stuff that they will retain as adults?

    Remember, IQ is an approximation of g, it is not a pure mesure of g. It includes some s, things like vocabulary that tutors may cram into the kid’s brain for a test, but which quickly evaporates. Most cases where IQ has been “raised” involve this sort of teaching to the test. So the question is whether non-g knowledge crammed into a kid’s head temporarily can help, and the answer is probably No. This sort of short-term rise in tested IQ (not g) is limited, and does not even affect tests like digit span memory.

  9. @Steve Sailer
    But literacy percentages seem like the kind of thing that differs radically across time and place. So the quantity and quality of schooling seems to matter whether a kid learns to read or not.

    > If you can boost IQ temporarily in kids, can you teach them more stuff that they will retain as adults?

    Yes, it’s interesting that the following appear to be true:
    * IQ in adulthood is primarily genetic, and not subject to improvements via childhood training.
    * Brain plasticity is higher in children, and allows easier or more complete acquisition of some types of skills.

    These two statements are in some ways contradictory. You would assume that the higher plasticity of brains in childhood _would_ allow some learning of things that would be useful in IQ tests. But perhaps brain plasticity is only useful for the kinds of skills that are irrelevant in g-loaded tests.

    Perhaps learning in the critical window is a source of some of the Flynn-effect, or differences between African and American black IQ results.

  10. @Steve Sailer
    But literacy percentages seem like the kind of thing that differs radically across time and place. So the quantity and quality of schooling seems to matter whether a kid learns to read or not.

    Wouldn’t a confounding factor be the need for literacy? A medieval peasant with a genius level IQ would never have a need to learn to read and write; the rudiments of farming and even possibly how to improve farming efficiency could be mastered without that fancy “book learning.” It’s hard for people who read and write a lot to imagine anyone with any significant cognitive ability living as an illiterate, but in some cultures it might be possible. Illiterate people sometimes have prodigious memory capacity–if you can write things down you just don’t need that.

  11. anon[393] • Disclaimer says:
    @Steve Sailer
    If you can boost IQ temporarily in kids, can you teach them more stuff that they will retain as adults?

    For example, literacy is a big deal. If you get your IQ boosted high enough that you learn how to read, you probably won't forget as an adult, and it's a lot better to go through life as a pretty dumb person who can read than a pretty dumb person who is illiterate.

    I taught my daughter her alphabet with those magnetic letters which was sort of reading as opposed to reciting alphabet at as soon as she could speak which was quite early. It was quite easy,, her mom and I are +2 sds what was interesting is after a few months of no lessons she forget them all again and i had to start over, again not too hard but she again forgot them without continuing training. cant say what would have happened if i was more diligent and kept increasing lessons without stopping, but I suspected I could have gone much farther but i also suspected before they are developed fully to whatever level you are training youre sort of writing on sand

  12. @Redneck farmer
    We've had EPDs (Expected Progeny Difference) for choosing which bull's semen to use for breeding for decades now. If humans are just another mammal........

    EPDs…

    A powerful tool. At cocktail parties, I used to excitedly explain how great they are, and was looking forward to the day when we could select dating partners for our kids based on the same techniques.

    I noticed people got uncomfortable, so I’ve dropped the subject. But thinking back to some pop culture references, it was more of a thing back in the day. On the Michael J Fox show Family Ties, he tries to get Mallory to date one of his clean cut friends by saying: “look at this guy; he is genetically incapable of making less than $50,000 a year.” In the Ian Anderson song Heavy Horses he croons: “let me find you a filly for your fine stallion seed; to keep… the old line… going.”

    Certainly don’t hear such musings in today’s pop world.

  13. Its either “B iff G” (behavior B is possible if and only if a specific genotype G is instantiated) or “if G, then necessarily B” (genotype G is a sufficient cause for behavior B). Both claims are false; genes are neither a sufficient or necessary cause for any behavior.

    And genes can’t cause cognition because there are no psychophysical laws. Cognition is intentional and therefore irreducible to brain states/structure, genes and physiology.

    • Replies: @phil
    There you go again, creating strawmen. Please learn more about statistics and probabilities. Multiple genes and environmental factors interact in somewhat unpredictable ways. Multiple genetic influences combined (see, e.g., genetic complex trait analysis) account for a substantial amount of the variance
    across individuals in behavioral traits, including cognitive ability.
    , @Joe Schmoe

    Both claims are false; genes are neither a sufficient or necessary cause for any behavior.
     
    Perhaps I misunderstand you, but here goes. Genes cause cognitive ability, which then allow for problem solving behaviors. Regardless of culture or education or location, all people need to solve problems. Some people really struggle to solve problems, but to others it comes naturally aka genetic. So, it seems it is both necessary and sufficient for problem solving ability and behavior.
    , @phil
    By your criterion, environmental factors also cannot cause cognition. Statistically, however, varying educational levels account for some of the variation in IQ scores.
    , @gcochran
    So, cognition should work just as well after six shots of vodka. Interesting. Yet I can remember having some trouble making a technical presentation after I'd had an unexpectedly large can of Foster's Lager over lunch.
  14. @Steve Sailer
    But literacy percentages seem like the kind of thing that differs radically across time and place. So the quantity and quality of schooling seems to matter whether a kid learns to read or not.

    literacy percentages seem like the kind of thing that differs radically across time and place. So the quantity and quality of schooling…

    Isn’t is axiomatic that anyone who can walk and talk, sans a disability, can learn to read and write if they deem it something important?

    My old school neighbor with a keen business sense runs a moderately successful enterprise, yet is essentially illiterate. He grew up where schooling was shunned: no books, no practice of writing to record events, etc. (Maybe it helped him hone his memory). Today’s world has made letters a much more salient thing, and his kids and grandkids are as literate as one would expect, and I’m sure he could have learned to read on his own if it was an important feature of his upbringing.

  15. @Steve Sailer
    If you can boost IQ temporarily in kids, can you teach them more stuff that they will retain as adults?

    For example, literacy is a big deal. If you get your IQ boosted high enough that you learn how to read, you probably won't forget as an adult, and it's a lot better to go through life as a pretty dumb person who can read than a pretty dumb person who is illiterate.

    Can read + has read = the real difference maker.

    If all one reads is Twitter and Potter that Can Read is wasted.

    Terrible thing!

  16. We already know for an incontrovertible fact that little black African girls are smarter than you are and that is all we need to know.

    • Replies: @Redneck farmer
    So you're saying African men are the continent's problem?
  17. What is the strange and counterintuitive here is that the polygenic scores (PS) can predict education achievement (EA) better than IQ. One would think that EA has lower heritability than IQ so why PS predicts ≈15% of EA variance and only ≈9% of IQ variance?

    9% is still far from the twin based heritability estimates of 50%-70%. They are using 100’s of 1000’s of SNPs already. So what next? A non additive nonlinear polygenic score which means that the exploratory sample to develop its betas will have to be really huge.

    • Replies: @res

    What is the strange and counterintuitive here is that the polygenic scores (PS) can predict education achievement (EA) better than IQ. One would think that EA has lower heritability than IQ so why PS predicts ≈15% of EA variance and only ≈9% of IQ variance?
     
    I think that is usually because EA data is more easily available than IQ data so sample sizes are usually larger for EA studies. Here are the numbers from the current paper:

    Increasing GWA sample sizes to 78,000 IQ2; 13 and then to 280,000 IQ3; 14 paid off in increasing predictive power of GPS from 1% to 3% to 4%. Here we present results for IQ3.

    Educational attainment has led the way in terms of increasing GWA sample size, from 125,000 in 2013 (EA115) to 294,000 in 2016 (EA216) to 1.1 million in 2018 (EA317). The growing sample sizes increased the predictive power of GPS from 2% to 3% to 12% of the variance in educational attainment
     
    And just after that they explicitly make the same point I made above:

    Because ‘years of education’ is obtained as a demographic marker in most GWA studies, it was possible to accumulate samples sizes with the necessary power to detect very small effect sizes. It is more difficult to obtain very large sample sizes for intelligence, which needs to be assessed with a psychometric test administered to each individual, whereas years of education can be captured with a single self-reported item.
     
    The current paper uses additional methods to use the EA3 and IQ3 to create even better predictors.

    The supplementary materials have more details. Table S1 showing the correlations between their four outcome variables (EA/IQ at ages 12/16) was interesting. Correlations from 0.52 - 0.6 except EA at ages 12 and 16 which was 0.77.
  18. @RaceRealist88
    Its either "B iff G” (behavior B is possible if and only if a specific genotype G is instantiated) or “if G, then necessarily B” (genotype G is a sufficient cause for behavior B). Both claims are false; genes are neither a sufficient or necessary cause for any behavior.

    And genes can't cause cognition because there are no psychophysical laws. Cognition is intentional and therefore irreducible to brain states/structure, genes and physiology.

    There you go again, creating strawmen. Please learn more about statistics and probabilities. Multiple genes and environmental factors interact in somewhat unpredictable ways. Multiple genetic influences combined (see, e.g., genetic complex trait analysis) account for a substantial amount of the variance
    across individuals in behavioral traits, including cognitive ability.

  19. @Peter Frost
    predictive power increases from age 12 to age 16

    This is the point I tried to make in my criticism of Chanda's article. Cognitive ability in children is malleable and is thus a poor predictor of adult intelligence. We can do a lot to increase IQ in children, but this increase tends to wash out during adolescence and is largely gone by one's early twenties.

    It would be interesting to see whether this "adolescence wash-out" differs between human populations, both in its timing and in its magnitude.

    Are we really talking IQ boost?

    If it doesn’t translate to being permanent, it is kind of like discounts at the store bringing future sales to the present but not increasing the total sales over time.

    What am I missing?

  20. @RaceRealist88
    Its either "B iff G” (behavior B is possible if and only if a specific genotype G is instantiated) or “if G, then necessarily B” (genotype G is a sufficient cause for behavior B). Both claims are false; genes are neither a sufficient or necessary cause for any behavior.

    And genes can't cause cognition because there are no psychophysical laws. Cognition is intentional and therefore irreducible to brain states/structure, genes and physiology.

    Both claims are false; genes are neither a sufficient or necessary cause for any behavior.

    Perhaps I misunderstand you, but here goes. Genes cause cognitive ability, which then allow for problem solving behaviors. Regardless of culture or education or location, all people need to solve problems. Some people really struggle to solve problems, but to others it comes naturally aka genetic. So, it seems it is both necessary and sufficient for problem solving ability and behavior.

  21. This must be one of the most interesting, challenging (very multidisciplinary), and exciting (in part due to the fact it is so controversial) fields to work in today.

    …to build prediction models of general cognitive ability and educational achievement.

    I think many things can affect achievement, eg motivation and opportunity (to name two of the obvious ones) — but it is measurable, and high cognitive ability is no doubt on average necessary for high achievement — can cognitive ability/potential, g, and its genetic foundations, be seen as a separate area of study?

  22. I honestly don’t know what the hell this research is trying to tell us. Are they saying that genes are good at predicting IQ and grades in schools, or they saying grades in school + IQ are better at predicting behavior than parental education level or occupation? Or are they saying something else entirely?

    Give me ambiguity or give me something else!

  23. @RaceRealist88
    Its either "B iff G” (behavior B is possible if and only if a specific genotype G is instantiated) or “if G, then necessarily B” (genotype G is a sufficient cause for behavior B). Both claims are false; genes are neither a sufficient or necessary cause for any behavior.

    And genes can't cause cognition because there are no psychophysical laws. Cognition is intentional and therefore irreducible to brain states/structure, genes and physiology.

    By your criterion, environmental factors also cannot cause cognition. Statistically, however, varying educational levels account for some of the variation in IQ scores.

  24. @obwandiyag
    We already know for an incontrovertible fact that little black African girls are smarter than you are and that is all we need to know.

    So you’re saying African men are the continent’s problem?

  25. @utu
    What is the strange and counterintuitive here is that the polygenic scores (PS) can predict education achievement (EA) better than IQ. One would think that EA has lower heritability than IQ so why PS predicts ≈15% of EA variance and only ≈9% of IQ variance?

    9% is still far from the twin based heritability estimates of 50%-70%. They are using 100's of 1000's of SNPs already. So what next? A non additive nonlinear polygenic score which means that the exploratory sample to develop its betas will have to be really huge.

    What is the strange and counterintuitive here is that the polygenic scores (PS) can predict education achievement (EA) better than IQ. One would think that EA has lower heritability than IQ so why PS predicts ≈15% of EA variance and only ≈9% of IQ variance?

    I think that is usually because EA data is more easily available than IQ data so sample sizes are usually larger for EA studies. Here are the numbers from the current paper:

    Increasing GWA sample sizes to 78,000 IQ2; 13 and then to 280,000 IQ3; 14 paid off in increasing predictive power of GPS from 1% to 3% to 4%. Here we present results for IQ3.

    Educational attainment has led the way in terms of increasing GWA sample size, from 125,000 in 2013 (EA115) to 294,000 in 2016 (EA216) to 1.1 million in 2018 (EA317). The growing sample sizes increased the predictive power of GPS from 2% to 3% to 12% of the variance in educational attainment

    And just after that they explicitly make the same point I made above:

    Because ‘years of education’ is obtained as a demographic marker in most GWA studies, it was possible to accumulate samples sizes with the necessary power to detect very small effect sizes. It is more difficult to obtain very large sample sizes for intelligence, which needs to be assessed with a psychometric test administered to each individual, whereas years of education can be captured with a single self-reported item.

    The current paper uses additional methods to use the EA3 and IQ3 to create even better predictors.

    The supplementary materials have more details. Table S1 showing the correlations between their four outcome variables (EA/IQ at ages 12/16) was interesting. Correlations from 0.52 – 0.6 except EA at ages 12 and 16 which was 0.77.

    • Replies: @utu
    Good try but not good enough. Educational achievement (EA) is supposed to be a function of IQ and environment. If they develop a predictor function that does well (15%) to predict EA this function is predicting well because it is predicting IQ part not the environment which suppose to be not correlated with genes. When the same function predicts only 9% of IQ then one may conclude that the 15% result is partly spurious or the environment affecting EA has a genetic component that is not correlated with IQ. This genetic component ended up in their polygenic score and is messing up correlation with IQ.
  26. @RaceRealist88
    Its either "B iff G” (behavior B is possible if and only if a specific genotype G is instantiated) or “if G, then necessarily B” (genotype G is a sufficient cause for behavior B). Both claims are false; genes are neither a sufficient or necessary cause for any behavior.

    And genes can't cause cognition because there are no psychophysical laws. Cognition is intentional and therefore irreducible to brain states/structure, genes and physiology.

    So, cognition should work just as well after six shots of vodka. Interesting. Yet I can remember having some trouble making a technical presentation after I’d had an unexpectedly large can of Foster’s Lager over lunch.

  27. @res

    What is the strange and counterintuitive here is that the polygenic scores (PS) can predict education achievement (EA) better than IQ. One would think that EA has lower heritability than IQ so why PS predicts ≈15% of EA variance and only ≈9% of IQ variance?
     
    I think that is usually because EA data is more easily available than IQ data so sample sizes are usually larger for EA studies. Here are the numbers from the current paper:

    Increasing GWA sample sizes to 78,000 IQ2; 13 and then to 280,000 IQ3; 14 paid off in increasing predictive power of GPS from 1% to 3% to 4%. Here we present results for IQ3.

    Educational attainment has led the way in terms of increasing GWA sample size, from 125,000 in 2013 (EA115) to 294,000 in 2016 (EA216) to 1.1 million in 2018 (EA317). The growing sample sizes increased the predictive power of GPS from 2% to 3% to 12% of the variance in educational attainment
     
    And just after that they explicitly make the same point I made above:

    Because ‘years of education’ is obtained as a demographic marker in most GWA studies, it was possible to accumulate samples sizes with the necessary power to detect very small effect sizes. It is more difficult to obtain very large sample sizes for intelligence, which needs to be assessed with a psychometric test administered to each individual, whereas years of education can be captured with a single self-reported item.
     
    The current paper uses additional methods to use the EA3 and IQ3 to create even better predictors.

    The supplementary materials have more details. Table S1 showing the correlations between their four outcome variables (EA/IQ at ages 12/16) was interesting. Correlations from 0.52 - 0.6 except EA at ages 12 and 16 which was 0.77.

    Good try but not good enough. Educational achievement (EA) is supposed to be a function of IQ and environment. If they develop a predictor function that does well (15%) to predict EA this function is predicting well because it is predicting IQ part not the environment which suppose to be not correlated with genes. When the same function predicts only 9% of IQ then one may conclude that the 15% result is partly spurious or the environment affecting EA has a genetic component that is not correlated with IQ. This genetic component ended up in their polygenic score and is messing up correlation with IQ.

    • Replies: @res

    Good try but not good enough.
     
    My primary point was the sample size difference. Which you completely ignored in your response.
  28. @utu
    Good try but not good enough. Educational achievement (EA) is supposed to be a function of IQ and environment. If they develop a predictor function that does well (15%) to predict EA this function is predicting well because it is predicting IQ part not the environment which suppose to be not correlated with genes. When the same function predicts only 9% of IQ then one may conclude that the 15% result is partly spurious or the environment affecting EA has a genetic component that is not correlated with IQ. This genetic component ended up in their polygenic score and is messing up correlation with IQ.

    Good try but not good enough.

    My primary point was the sample size difference. Which you completely ignored in your response.

    • Replies: @utu
    I did not have to. Because if the PS is developed for EA on large sample and it does not correlate as well with IQ as with EA then because EA=k*IQ+Envir then it implies that the PS picked up signal that correlates with Envir which must be spurious unless the Envir has a genetic component that is not correlated with IQ.
  29. @res

    Good try but not good enough.
     
    My primary point was the sample size difference. Which you completely ignored in your response.

    I did not have to. Because if the PS is developed for EA on large sample and it does not correlate as well with IQ as with EA then because EA=k*IQ+Envir then it implies that the PS picked up signal that correlates with Envir which must be spurious unless the Envir has a genetic component that is not correlated with IQ.

Comments are closed.

Subscribe to All Steve Sailer Comments via RSS
PastClassics
The unspoken statistical reality of urban crime over the last quarter century.
Which superpower is more threatened by its “extractive elites”?
How a Young Syndicate Lawyer from Chicago Earned a Fortune Looting the Property of the Japanese-Americans, then Lived...
Becker update V1.3.2