design, we over-simulated to generate 500 subjects with liability at or below the 10th percentile and 500 subjects above the 60th percentile. In the random design, we simulated 1000 trios irrespective of outcome and environmental stratum. We selected parameter values for genetic and environmental effects similar to those seen for common behavioral outcomes. However, since the goal of this work is to examine the comparative power of these approaches, the values of the parameters is not of critical importance since the relative power of the approaches is unlikely to change under different effect sizes. We assumed a minor allele frequency of the protective allele of 0.3, an overall trait heritability of 0.50, the variance accounted for by the family environment of 0.25. We varied the impact of the protective SNP and individual environment. For each of 1000 iterations under each study design, we tested the association between the offspring protective SNP and protective individual environment with outcome via logistic regression. Power was calculated at α =.05 or the proportion of Monte Carlo replicates that yielded a p-value less than .05.
