To illustrate the difference in power between TATES and Simes, we simulated phenotypic data according to 1-factor Rasch models, with factor loadings of .8660, .9220, or .9747, indicating correlations of .75, .85 and .95 between the phenotypes, respectively. The GV effect was modeled either on the latent factor (like Figure 1a; Tables S13, S14, S15), or directly on one of the 20 phenotypes (like Figure 1e; Tables S16, S17, S18).
