The success in eliminating the bias of imputing SNPs across arrays by using the intersection approach brought to the fore practical considerations of when use of public controls genotyped on multiple arrays is worthwhile. Thus, in a final set of analyses, we examined the effects of increasing sample size and reducing imputation accuracy on statistical power when using public controls under two simplified scenarios. First, adding public controls to an existing genotyped sample of cases and controls meaningfully increased power with the addition of as little as one-third of the original control sample if imputation accuracy remained moderate to good (average R2 ≥ 0.7). Second, designing a GWAS to rely on public controls to supplement or replace study controls showed even more marked increases in statistical power by focusing fixed resources on increasing the number of cases genotyped as well as boosting the size of the control group.
