The weighted-sum method is compared to the CAST, CMC, and variant-by-variant methods, which were discussed in the introduction and are described in more detail in Comparison with other Methods. For each set of parameters, 100 datasets are simulated, the four methods are applied, and the proportions of significant outcomes are used as the power estimates. To mimic a genome wide study of about 20,000 fairly independent human genes, we calculate a p-value for each gene, and use a significance threshold of 0.05/20000 = 2.5×10−6 in all power simulations.
