It is possible that both age and sex replication rates are downwardly biased due to differences in age and sex distributions (Table 1). To quantify the potential effects of heterogeneous sample sizes and unbalanced study designs, we performed resampling studies within the UC discovery cohort. Demographic effect replication rates were recalculated using 60 samples that were race, sex, and age (+/−3years) matched to the UW distribution (Figure 1B, 1E; see Methods). We found that 34% of sex effects and 15% of age effects replicated by simulation, supporting the conclusion that sample size and demographic heterogeneity do generate significant covariate associations that our replication studies are unable or underpowered to detect.
