Simulation results for using different weights to construct SNP set tests.

<p>(A) QQ plot of -log10 p values from SNP set tests using different SNP weights under the null simulations. Tests using different weights all control type I error well. (B) Power to detect causal blocks by SNP set tests using different SNP weights in the simulation setting where <b>α</b> = <b>(0.4, 0.4)</b>. (C) Power to detect causal blocks by SNP set tests using different SNP weights in the simulation setting where <b>α</b> = <b>(0.4, 0).</b> For both (B) and (C), Power are evaluated at a genome-wide significance threshold of 1x10^-4. Standard errors are computed across 1,000 simulation replicates. The x-axis shows the proportion of causal SNPs that have identical values for the two annotations, which measures correlation between the two annotations.