It is often the case that when people do GWAS studies they quantify the performance of their methods based on localization and power test. Localization test is computing the distance between the true causal snp and the most significant snp across multiple replicates. In order to do power analysis, people compare the most significant snp in each replicate to the null distribution. I have simulated data and a quantitative trait for each sample and was wondering how would I possibly setup a power test. In other words, how would i generate this null distribution?