No, you wouldn’t, because intelligence is causally driving outcomes of social status, income, and education. (Think about what it would mean to compare two populations with different genetic potential but who still somehow wind up with identical income & education...) Like the fallacy of controlling for intermediate variables, controlling for outcomes is controlling away the effect. It would be like running a drug trial in which you controlled for deaths.
If you are, for some improbable reason, deeply concerned that your genetic correlates are some sort of very subtle population stratification that your PCA missed, you can check using a within-family design, which by construction keeps many variables constant without illicitly controlling for outcome variables; and we already know that the IQ hits survive this test because Rietveld et al 2013 checked the original hits (“The polygenic score remains associated with educational attainment and cognitive function in within-family analyses (table S25)”), “Polygenic Influence on Educational Attainment: New Evidence From the National Longitudinal Study of Adolescent to Adult Health” Domingue et al 2015 replicated it in the USA, and the still-upcoming SSGAC paper also finds no residual confounding.
How do you handle a scenario where the causal arrow goes both ways, i.e. intelligence drives employability and wealth, while nutrition and prenatal care drive intelligence?
You might handle it with a longitudinal SEM or causal net since you have time separating effects (parental intelligence comes before wealth which comes before nutrition/prenatal-care which comes before childrens’ intelligence); but for that specific case, nutrition & prenatal care are already largely ruled out as causally relevant since they fall under ‘shared environment’, which for IQ in the West is very low.
You’d need to compare populations of similar social status, income, and education for the difference to be meaningful.
No, you wouldn’t, because intelligence is causally driving outcomes of social status, income, and education. (Think about what it would mean to compare two populations with different genetic potential but who still somehow wind up with identical income & education...) Like the fallacy of controlling for intermediate variables, controlling for outcomes is controlling away the effect. It would be like running a drug trial in which you controlled for deaths.
If you are, for some improbable reason, deeply concerned that your genetic correlates are some sort of very subtle population stratification that your PCA missed, you can check using a within-family design, which by construction keeps many variables constant without illicitly controlling for outcome variables; and we already know that the IQ hits survive this test because Rietveld et al 2013 checked the original hits (“The polygenic score remains associated with educational attainment and cognitive function in within-family analyses (table S25)”), “Polygenic Influence on Educational Attainment: New Evidence From the National Longitudinal Study of Adolescent to Adult Health” Domingue et al 2015 replicated it in the USA, and the still-upcoming SSGAC paper also finds no residual confounding.
How do you handle a scenario where the causal arrow goes both ways, i.e. intelligence drives employability and wealth, while nutrition and prenatal care drive intelligence?
You might handle it with a longitudinal SEM or causal net since you have time separating effects (parental intelligence comes before wealth which comes before nutrition/prenatal-care which comes before childrens’ intelligence); but for that specific case, nutrition & prenatal care are already largely ruled out as causally relevant since they fall under ‘shared environment’, which for IQ in the West is very low.