JayMan471

JayMan · @JayMan471

27th Sep 2014 from TwitLonger

My response to yet another challenge to behavioral genetics


OK David Colquhoun, first of all, let me says thanks boatloads for giving me the paper I have been tearing up all of Twitter to find. Now, my commentary. First, I know you don't accept either heredity or IQ despite overwhelming evidence for both. Your model seems to be, as based on this paper, that many health and behavioral outcomes have a large random component. On that point, you're quite right randomness is important, but even randomness is NOT purely random, but itself is quite heritable http://jaymans.wordpress.com/2014/03/04/environmental-hereditarianism/.

However, behavioral genetic studies set substantial limits to the importance to pure randomness to outcomes. We know that many health outcomes and behavioral traits are substantially heritable. The existence of genetic differences between MZ twins – which do indeed exist – only serves to lead twin studies to underestimate heritability (though perhaps not by much). And yes, you're quite right, even very tiny genetic differences can lead to huge phenotypic differences. But, again, see the previous point. Heritability estimates (H^2) are surely not *overestimated* by twin studies, as the results come not just from MZT-DZT studies, but, from MZ apart, adoption studies, and direct genomic comparisons (GCTA) looking within and between families.

Nor do these studies underestimate the effect of shared environment (c^2), because these methods all produce zero shared environment. As well, it's not lack of statistical power either, as large meta-analyses (often with total N > 30,000) also find c^2 = 0. See http://jaymans.wordpress.com/2014/03/31/the-son-becomes-the-father/ &
http://jaymans.wordpress.com/2014/04/15/more-behavioral-genetic-facts/

That's not even to mention the recent Swedish population studies that confirm lack of most suspected environmental effects.

As your paper points out, much of what is attributed to unshared/unique environment (e^2) is actually measurement error. Studies that correct for error find higher H^2. So that sets an upper bound to the effect of randomness on population variance.

Where you have bona fide low heritability, also lack of c^2, as in the case of cancer (and for that matter, homosexuality), but a nonetheless real condition, that strongly suggests one class of agents are responsible: pathogens, as was argued by Greg Cochran and Paul Ewald: http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.182.5521&rep=rep1&type=pdf

Reply · Report Post