Sex differences and sex × strain interactions are found for a majority of the behavioral traits in this study, and as reported in previous studies (Chesler et al. 2002a; Valdar et al. 2006), sex interactions with genotype are typically more profound than main effects of sex. When significant sex differences are likely to be present, there are several approaches for considering them in further genetic analyses. Male and female phenotypic data can be analyzed separately for QTL mapping and genetic correlations. This is an approach most practically employed in tools such as GeneNetwork, which make use of a single mean value for each strain. Another approach which generates a single-trait value per strain is to regress male and female data (Fernandez et al. 1999). Finally, a nested analysis can be employed at each locus to appropriately consider the replication within strain and across sex within each genotype class. The latter approach reduces multiple testing concerns while retaining maximal statistical power, but requires complex permutation approaches (Peirce et al. 2008). By adopting a liberal bias in the detection and treatment of
