Fig. 2From: Combining controls can improve power in two-stage association studiesPower of both methods is equivocal at high effect sizes (high absolute log odds ratios) and at low effect sizes (log odds ratio near zero). The main region in which power can differ is at moderate effect sizes. A good metric for difference in power is the average difference in power (marked ‘i’). The maximum difference in power (marked ‘ii’) is also considered. This plot shows analytic rather than simulated resultsBack to article page