Interest in testing for gene-environment interactions in relation to common, complex disorders is widespread and growing. A PubMed search using the term “gene-environment interaction” produces more than 1,300 papers, more than half of which were published within the last five years. Across multiple disciplines, gene-environment interaction is an active area of research, with large literatures testing candidate gene-environment interactions (e.g., Caspi et al. (2003); Bakermans-Kranenburg and van Ijzendoorn (2006); Petersen et al. (2012); Duncan and Keller (2011), Widaman et al. (2012), Keller (2014)), and growing interest in testing for gene-environment interaction in the context of genome-wide association studies (GWAS) (Khoury and Wacholder (2009); Murcray et al. (2009)). Characterizing genotype by environment interactions presents many challenges, including questions about how best to measure the environment, the relative importance of the timing and duration of the environmental exposure, and the potential for increased problems with multiple testing when large-scale genotypic information is crossed with many potential environmental variables of interest. However, one area that has received little scrutiny is the statistical methods that are routinely applied to test for gene-environment interaction. The
