In the second day 3 video, it mentions several assumptions of the classical twin design (random mating, no GE correlation, no GxE interaction) that could be difficult to test.
If any of these assumption violations are present, what do they do to the model?
Are there generalizations that can be made like “if GxE interaction is present, it will (inflate/deflate) estimates for specific variance components”?
Yes!
(Positive) assortative mating on a particular trait will act to make DZ twins more similar phenotypically for that trait than what they would be under random mating. Consequently this will inflate the DZ correlation and show up as inflated “C” in the classical twin model.
From memory (and Im sure other faculty will correct me if wrong!):
AxE goes into E
AxC goes into A
AE correlation goes into A
AC correlation goes into C
For further discussion of how AxC, AxE and gene-environment correlation affect estimates from the classical twin design c.f.:
Variance components models for gene-environment interaction in twin analysis - PubMed
Dave
One addition to ‘(Positive) assortative mating on a particular trait will act to make DZ twins more similar phenotypically for that trait than what they would be under random mating. Consequently this will inflate the DZ correlation and show up as inflated “C” in the classical twin model.’ If you fit an AE model though, the “A” estimate would be inflated.