As the original Simes procedure does not take into account the correlations among the p values (originating in the phenotypic correlations), TATES is expected to increasingly outperform Simes as the phenotypic correlations increase. Given low to modest phenotypic correlations, the gain in power acquired with TATES varies from low (1%) to modest (9%) (the latter observed in a 4-factor model with a phenotype-specific GV effect; Table S12). Additional simulations (Tables S13, S14, S15, S16, S17, S18), however, show that, as phenotypic intercorrelations increase in magnitude (.75, .85, .95), the power of TATES can be as much as 10%–19% higher than the power of the Simes procedure, with TATES especially being more powerful when the GV effect is specific to one of multiple correlated phenotype. As TATES is comparable to Simes in computational effort, phenotypes within a trait are almost invariably correlated, and the location of the GV effect is generally unknown (i.e., could be phenotype-specific), one is well-advised to adopt TATES.
