We return below to consideration of results for studies in the Yoruban population. Focussing now on the power curves in the top row of Figure 2 several features are evident. The first is the profound effect of sample size. Effect sizes of 1.5 or smaller might be typical of what would now be expected for most variants affecting susceptibility to common human diseases [5]. For effect sizes at the top of this range (1.3–1.5) very large studies (say 2,000–3,000 cases and the same number of controls) are needed to have reasonable power, while for smaller effect sizes even studies of 5000 cases and 5000 controls have very little power. This ties in with growing empirical evidence. For example, for Crohn's disease, the WTCCC study, of 2000 cases and 3000 controls found 9 loci with p<5×10−7, whereas several smaller studies published around the same time each found only one or two of the loci, with little overlap across these smaller studies, consistent with each having modest power for the larger set of loci. Further, recent meta-analyses of 4,539 cases for type
