Investigators should demonstrate that the sample being employed has adequate power to detect an interaction effect for the variables under study. Power computations should be presented regardless of the nature of the finding. In other words, studies with positive findings should be encouraged to present power computations as well (K. S. Button et al., 2013). Computations should be specific to the statistical technique, distributions of the variables under investigation, and the hypothesized form of the interaction. Power analyses should assume realistic cGxE effect sizes. GWAS suggest that main effects are typically of a small magnitude, most accounting for less than 1% of the variance in a psychiatric phenotype. If the investigator has reason to believe that a larger effect size is likely for their study this should be clearly spelled out and justified. For instance, one might hypothesize that genes exert stronger effects on endophenotypes (e.g., neuroimaging outcomes) that are, arguably, more proximal to their action (though see Flint & Munafo, 2007; Munafo & Flint, 2009). Or, it is possible that the use of a phenotypic measure thought to be of much greater reliability and validity than existing clinical phenotypes could enhance power. Finally, investigators should use
