Emil O W Kirkegaard

The sibling correlation for IQ is about 0.40, i.e. most population variance is within families. Unless Richardson thinks that siblings raised together have vastly different social class backgrounds, his theory is... problematic.

This is not very productive use of my time, but oh well. Consider the possibility that I don't argue in bad faith & you just don't understand the issues at hand. Sauce & Matzel (S&M) claim that GWA research on intelligence is non-replicable, citing Chabris (2012) & Manuck (2014)

In fact both Chabris and Manuck say that _candidate genes_ are non-replicable. Chabris's paper doesn't consider GWAS while Manuck says "GWA studies have found novel loci... related to many complex physical traits and disorders, often well replicated and sustained on

meta-analytic review". Similarly, S&M say that the genetic markers used in Harlaan et al. (2005) "were obtained by standard GWAS techniques". In reality, they were candidate genes and the first GWAS hits on IQ weren't found until years later. S&M don't know what GWAS (or GCTA or

LD score regression etc.) means and thus fail to understand the developments in IQ genomics in recent years. I don't know what they mean by "functional units in an evolutionary sense" or why the number of causal variants should be small rather than large. Note that much of the

genetic variance in IQ is probably "sand in the machine", i.e. small deleterious mutations. Re their treatment of the Flynn effect (and test scores in general), treating raw score gains as gains in intelligence is a category error: https://www.gwern.net/docs/iq/2004-wicherts.pdf … Re SDs & environment,

the variance components in a behavior genetic model like the ACE are obtained by squaring the standardized regression weights that indicate the effects of the latent variables (A, C, E) on the phenotype. Therefore, if c2=20%, the standardized weight of C on IQ is √.2≈.45.