Two remaining points have to do with the aforementioned issues of statistical power and multiple testing. Statistical power should be the foremost consideration of any proposed genetic association study, with sample sizes assembled accordingly. Although power derived from large samples is necessary for the analysis of complete genome-wide data because of the number of individual tests performed, it is also worth noting that many studies of candidate genes are also underpowered to detect effects of the sizes currently expected on the basis of GWAS data from other behavioral or psychiatric phenotypes (allelic odds ratios [OR] in the range of 1.1–1.15). As such, social scientists should be aware that statistical power in genetic association studies is a function of several elements of the research design including, but not limited to, the coding of phenotype (e.g., case–control versus continuous outcome) and genotype (e.g., additive versus dominant or recessive modes of inheritance), minor allele frequency (MAF), type I error rate, and the hypothesis under consideration (i.e., genetic main effect, gene–environment interaction, or even gene–gene interaction). To illustrate, let us consider how several of
