Uncle Judea, Melanin, Genetics, and Educational Attainment...
Highly Recommended: Judea Pearl and Dana Mackenzie: The Book of Why: The New Science of Cause and Effect https://books.google.com/books?isbn=0465097618. And: Aha! So I am not as stupid as I am ugly after all!:
Sokrates: "Your intuition is exactly right: SNPs for (or chromosomally linked to) melanin would predict educational attainment, in a sample drawn from the current over-all American population. The R^2 just for them might not be 10%, but it would certainly be non-negligible...
...Indeed, I would be somewhat surprised melanin-linked SNPs wouldn't be predictive even in a sub-sample of just those categorized as "black". The same would be true, though I suspect to a lesser extent, for SNPs (linked to) curly hair. The more fundamental point is that social and cultural inheritance, together with endogamy, mean that that there are certainly SNPs which predict your class background and the cultural traditions you were exposed to. (If we haven't identified SNPs which distinguish Baptists from Congregationalists among current Americans, it's pretty certainly because we've just not looked for them.)
You would expect a non-trivial ability to predict educational attainment and cognitive function (bracketing, for now, the extent to which we can measure "cognitive function" independently of "educational attainment") from a GWAS even in a blank-slate world, where all differences in mental function were purely a result of learning and experience. (For the record, of course we don't live in a blank-slate world.) This is one of those situations where Uncle Judea would, correctly, tell us to draw the DAG, and doing so would show an un-blocked back-door path: (ego's genes) <- (ego's parents' genes) <-> (ego's parents' environment) -> (ego's environment) -> (ego's IQ), which we don't really know how to deal with.
The new Nature Genetics paper doesn't, on first scan, show any awareness of this sort of issue at all...
DeLong: Am I profoundly stupid, or is Uncle Judea's framework of causal confounders—colliders—mediators a huge advance, perhaps not in helping those of you who think carefully do non-stupid statistics, but in helping those of us who do not think carefully do non-stupid statistics, and in providing a royal road to teaching people how to do not-stupid statistics?
Sokrates: "No, it really is that awesome!...
...There are other people who can claim a great deal of shared credit for that advance, especially in philosophy. But it really is that important, for all the reasons you give. I'd say it's helpful to those do think carefully, as well!
That said, Pearl's technical books are not easy to learn from.
I do think there's something to the expression of decision theory in terms of bets and money that isn't handled by just observing "well, that was the first technological context where we really needed to understand all this". The relevant analogy would be if we had formulated the laws of Newtonian physics in terms which only made sense for 17th century clockwork, so that before you could try to study, say, cloud formation or polymers that you had to pretend everything was really a rigid body.
