Erroneous Monk

First, on skin color: the authors admit to poor "tagging" on the color question in their data sample, reducing their phenotype sample to ~70% of full. Further, their reported correlation between color and ancestry is on the low end of previous research (0.39 vs 0.5).

1) A lower-than-average correlation implies that the lack of significance in the skin color moderator may simply be either sampling or error.

With that said, the sample size is really large for this type of study (7,000) so I doubt it's a sampling issue. But given that they note a prevalence of phenotype entry errors, I'm leary of taking that result too seriously.

I'm a novice on the genetics, but I'm extremely skeptical of their methodology here as well. Their genetic predictors (eduPGS) have only half the predictive ability for African Americans as they do for Europeans.

That's to be expected, as most of that research is done on European populations. But we should step back and consider the causal theory: if Euro ancestry predicts g through genetic inheritance, then that should be expressed primarily through those genes.

None of that makes these results invalid. And eduPGS doesn't predict all cognitive variability even among Europeans. It's intrinsic to the model that there would be leftover unmeasured genes contained in Euro ancestry.

But the lack of portability of these eduPGS scores is worrying for taking these results seriously. A lot more research on Africans and eduPGS measures is sorely needed.

To switch gears, you could also argue that the paper is an UNDERestimate of genetic factors. In controlling for parental education, they're arguably filtering out a huge portion of genetic influence.

And finally, this is predictive, not causal research. It could be the case that African ancestry predicts, say, exposure to lead, driving the other results. There are a whole host of unexamined environmental factors that could confound this analysis. Including parental wealth.