Evidence for sex-differences in the genetic architecture of complex human traits was found for 6 of the 122 variables (in 4 traits the difference was explained by different genes in men and women and in 2 traits by different environmental influences in men and women). The observed number is in line with that expected if the type-I error rate is 5%, indicating that in our data sex-differences in the genetic architecture of complex phenotypes are rare. Power analyses (see Table S6) indicated that power in our samples is sufficient to detect sex by genotype interactions for nearly all traits. To attain sufficient power to detect differences between DZss and DZos correlations (based on likelihood-ratio tests) sample sizes need to be larger as the correlations between relatives decrease. The smallest samples sizes in this study were for measures of brain function, which tend to show high heritability and high correlations in first-degree relatives, while the low DZ correlations were for traits such as birth weight for which sample sizes were large.
