The results are presented in Table 1 corresponding directly to the models in Figure 2. The impact of the protective environment was varied at R2 of .0225, .04, and .0625. The effect of the protective allele was varied at effect sizes, indexed as deviation from the heterozygote mean of −.15, −.20 and −.25, yielding an additive genetic variance of .009, .017 and .026 respectively at a minor (protective) allele frequency of .3. Power is presented at an alpha of .05. It is important to note that relative, and not absolute, power is of interest in comparative analyses such as this. It is expected that any factors which would increase or decrease power, such as change in effect size or sample size, would similarly affect the power of all approaches. As expected, case-control sampling yields substantial improvements over random sampling for detecting protective SNP and environment effects across all effect sizes. The low risk design, where subjects are ascertained at a low liability, yields power equivalent to a case-control design to detect SNP effects, and much higher power to detect the
