where yijis the phenotype being measured for Straini and Sexj. In addition to testing these effects, we estimated the magnitude of strain, sex and strain × sex effects by estimating the intra-class correlation, a partial ω2, for each effect per trait. Partial ω2 was estimated as the proportion of variance accounted for by the main or interaction effects relative to the total phenotypic variance. All variance components were estimated using the REML option of SAS PROC VARCOMP. This method can be biased due to departures from normality, a common phenomenon for behavioral traits. The percent variance accounted for by strain is considered by some to be an estimate of broad-sense heritability (Hirsch 1967; Lynch & Walsh 1998) for clones, and is formulated as the strain intra-class correlation: (2)
