I am wondering if too many define variables will decrease E and increase A?
For instance, I want to estimate the heritability of coronary heart disease (CHD), should I adjust all the risk factors of CHD as much as possible, such as smoking, drinking and BMI? I'm afraid that it will overestimate A and underestimate E since I saw a revelent paper.
Many thanks!
Printer-friendly version
By "define variables", I assume you mean covariates in the model for the phenotypic means (right?), as that is the most common use of definition variables in BG models. But, what do you mean by "A" and "E" here? If you mean the raw additive-genetic and nonshared-environmental variance components, then yes, at least one of them would be expected to decrease, unless your set of definition variables has an R² of exactly zero with the phenotype in your dataset. But if you mean the standardized additive-genetic and nonshared-environmental variance components, then that's a tougher question to answer. It would depend on the extent to which each of your definition variables partials out additive-genetic variance vis-a-vis nonshared-environmental variance in the phenotype, and on the correlations among the definition variables.
I lack subject-matter expertise in the genetics of CHD, and I haven't (yet) read the paper you linked, so I cannot presently offer any trait-specific advice.
Thanks for your explanation!