Links for February ’22

Genetics

  • The “Golden Age” of Behavior Genetics? by Evan Charney. The author is a political scientist best known for his anti-hereditarian screeds, of which this is the latest. He likes to discuss various random phenomena from molecular biology, describing them as inscrutably complex, a hopeless tangle in the face of which genetic analyses are futile. Unfortunately, his understanding of the statistical models designed to cut through that tangle is very limited. For example, he endorses Eric Turkheimer’s howler that genome-wide association studies are “p-hacking”, and makes a ridiculous argument about GWAS findings being non-replicable (p. 8)–he does not appear to know, among other things, that statistical power is proportional to sample size (Ns in the studies he cites range from ~100k to ~1000k), that the p-value is a random variable, or that SNP “hits” are subject to the winner’s curse (he cites but evidently never read Okbay et al., 2016 and Lee et al., 2018, wherein it is shown that GWAS replication rates match theoretical expectations). He seeks to identify and amplify all possible sources of bias that could inflate genetic estimates, while ignoring biases in the opposite direction (e.g., the attenuating effect of assortative mating on within-sibship genomic estimates). Often the article is weirdly disjointed, e.g., Charney first discusses how sibling models have been used to control for population stratification, and then a couple of pages later says that it is impossible to know whether differences in religious affiliation are due to heritability or stratification. All in all, the article is a good example of what Thomas Bouchard has called pseudo-analysis.
  • Neither nature nor nurture: Using extended pedigree data to elucidate the origins of indirect genetic effects on offspring educational outcomes by Nivard et al. Contra naysayers like Charney, we are in the midst of a genuine golden age in behavior genetics. The underlying reason is the abundance of genomic data, which has spurred the development of so many new methods that it is hard to keep up with them. This preprint is the latest salvo in the debate about indirect genetic effects. Previous research has found indirect parental genetic effects in models where child phenotypes are regressed on child and parent polygenic scores. This study refines the design by doing the regression on adjusted parental polygenic scores that capture the personal deviations of parents’ scores from the mean scores of the parents and their own siblings. This refined design finds scant evidence for indirect parental genetic effects on children’s test scores, suggesting instead that apparent indirect effects are grandparental or “dynastic” effects of some sort. I think assortative mating is the most likely culprit. A limitation of this study is that even with a big sibling sample, the power to discriminate between different models is not high. Moreover, the study does not actually test the difference between the βs of the sibship and personal polygenic scores, and instead reasons from differences in significance, which is bad form.
  • The genetics of specific cognitive abilities by Procopio et al. This impressively large meta-analysis finds the heritability of specific abilities to be similar to that of g. That may be the case although most measures of non-g abilities in the analysis are confounded by g. They can formally separate g and non-g only in the TEDS cohort which has psychometrically rather weak measures of g.

IQ

  • Ian Deary and Robert Sternberg answer five self-inflicted questions about human intelligence. The two interlocutors in this discussion are mismatched: Deary is the most important intelligence researcher of his generation, known for his careful, wide-ranging empirical work, while Sternberg is one of the greatest blowhards and empty suits in psychology, known for generating mountains of repetitive, grandiose verbiage and for his disdain for anything but the most perfunctory tests of the theoretical entities that proliferate in his writings. Sternberg’s entries provide little insight, but there is some comedy in first reading his bloviations and then Deary’s courteous but often quietly savage responses. Deary emphasizes the value of establishing empirical regularities before or even instead of formulating psychological theories; notes the ubiquity of the jangle fallacy in cognitive research; and argues that cognitive psychological approaches have not generated any reductionist traction in explaining intelligence. According to Deary, a hard problem in intelligence research is one of public relations, that is, getting “across all the empirical regularities known about intelligence test scores”, the establishment of which has been “a success without equal in psychology.”
  • More articles by Stephen Breuning that need retraction by Russell Warne. Stephen Breuning is an erstwhile academic psychologist who was caught fabricating data on a large scale. He received the first US criminal conviction for research fraud in 1988. Nevertheless, many of his publications have not been retracted and continue to be cited, e.g., in the influential meta-analysis of the effects of test motivation on IQ by Duckworth et al. (2011). Warne reviews four of Breuning’s unretracted studies and identifies a number of inconsistencies and implausibilities that point to data fraud. It might be useful to further analyze these studies with GRIM, SPRITE, and the like.

Sex and race

  • Sex differences in adolescents’ occupational aspirations: Variations across time and place by Stoet & Geary. More evidence for the gender equality paradox which postulates that sex differences are larger in wealthier and freer societies because heritable sex differences are suppressed in poorer societies where individuals have less choice.
  • Why Are Racial Problems in the United States So Intractable? by Joseph Heath. Most modern societies have dealt with ethnic and racial diversity either by trying to integrate minorities to the majority population, or by recognizing the separateness of minorities and devolving political power to them. Some countries have judged the success of these efforts through the lens of equal opportunity, while others have sought outcome equality. Heath argues that race problems involving African-Americans are so intractable and acrimonious because there is no agreement on whether integration or separatism should be pursued, nor on how success and failure in racial affairs are to be judged. He manages to squeeze a good deal of analytic juice from this simple model while avoiding “bad actor” explanations which attribute all racial problems either to white malevolence or black incompetence. Money quote: “[T]he best way to describe the current American ap­proach to racial inclusion would to be to say that it is attempting to achieve Singaporean outcomes using Canadian methods and legal frameworks.”


8 Comments

  1. RandomPerson

    Hello, I have sent you an email recently regarding getting in contact with you on Twitter. Could you please take the time to respond? I have some questions about some of your previous data, which is believed to be faulty by some people I’ve talked to recently. Thank you.

    • Dalliard

      Is this for me or Chuck? Only I use the Twitter account now, Chuck got a little too carried away with it.

      • RandomPerson

        Well, I was hoping to reach Chuck but I won’t have issues talking to you if you’re just as knowledgeable as he is regarding IQ-related issues.

        • Dalliard

          Sure, shoot away.

          • RandomPerson

            So, my first, very basic question is: in this link here (https://analyseeconomique.wordpress.com/2013/05/11/black-white-iq-gap-in-the-nlsy97-does-education-matter/) where Chuck conducted whatever he did, why is the sample size for AA subjects much larger than complete data for parent education in NLSY97? Also, The averages for ASVAB are also very strange, because he did not remove the dummy variables from ASVAB scores and they’re lowering average scores by bracket. I’m inclined to believe this since there’s a note for recoding EDU variables, but none for ASVAB. This is even more troubling when at lower brackets the STD is so close to the average. Why is that?

          • Dalliard

            That analysis is by Meng Hu, not Chuck. He used to write for this blog.

            The analysis is not very sophisticated (as MH would no doubt agree today) but it doesn’t seem obviously wrong to me at a glance. The sample sizes may appear to be inconsistent because survey weights were used to correct for sampling biases. Using weights will generally lead to less biased estimates of parameters like means and standard deviations, but because some observations receive a higher weight and others a lower weight, the descriptive statistics of the weighted sample will differ from the original sample.

            The means and SDs of the ASVAB scores seem plausible enough to me. Can you be more specific about your concerns and the reasoning behing them?

        • Meng Hu

          Hi, so I happened to read the comment section here, and found someone mentioning my old blog post. As Dalliard pointed out, this is not an anaysis I will rely on if I were to publish on a journal. You should check longitudinal studies of education effect on g vs non-g over time.

          As for my article, I looked at the B-W gap at different levels of parent education (which is why I “did not remove the dummy variables from ASVAB scores”). And I do not see where you would see the AA sample being larger than the total sample for parent education.

          Finally, the linked blog is dead, I can’t access anymore for years now. The only one blog I use is this one : https://menghublog.wordpress.com/

          Barely active but still there. I have some more refined analyses to come, but will post those on HV blog instead. I stayed in the dark for a while (i.e., reading but saying/doing nothing), but I will try to post more.

          In any case, it’s good to see you Dalliard keeping this blog going. As it’s still my favorite one. So, a big thank you for your dedication.

          • Dalliard

            Thanks and it’s great if you’re getting back into blogging.

Leave a Reply

Your email address will not be published.

© 2022 Human Varieties

Theme by Anders NorenUp ↑