Case-control studies

Following [36], we sampled 3000 cases and 3000 controls from each simulated population. Cases were randomly sampled from the upper 15% of phenotypic values in the population, and controls were randomly sampled from within 0.5 standard deviations of the population mean(as in [36]). This is the liability scale model (see [29]). We define a “GWAS” to be a study including all markers with MAF ≥5% and a re-sequencing study to include all markers. In all cases we used a minor allele count logistic regression as the single marker test. For single marker tests, the p-value cut off for significance is p ≤ 1e − 08 which is common in current GWAS [62, 98]. Power is determined by the percentage of simulation replicates in which at least one marker reaches genome wide significance.