Fig. 3.

Powers of the T statistic and other sweep detection methods—H12, nS_L, SweepFinder2, and Trendsetter—at the 1% FPR to detect soft selective sweeps from selection on standing variation on ν = 4 distinct sweeping haplotypes beginning at times $t\in \{200,500,1,000,1,500,2,000,2,500,3,000,3,500,4,000\}$ generations prior to sampling, for the European CEU (top) and sub-Saharan African YRI (bottom) human demographic models inferred with smc++. Analysis data consisted of phased haplotypes of length 1 Mb, with 1,000 replicates for each distinct scenario. Selective sweeps were simulated for five ranges of selection coefficients (s) spanning very weak to very strong, with s for each replicate drawn uniformly at random (from a log scale for s drawn across orders of magnitude). All inferences used a spectrum of K = 10 for likelihood computations.