Variance in fits of an exhaustive collection of 32,745 simple admixture graphs (five groups and two admixture events) resulting from ascertainment or random site subsampling expressed as squared Pearson correlation coefficient (<i>R</i>^<i>2</i>).

Correlation admixture graph log-likelihood scores (LL) or worst f4-statistic residuals (WR) was explored for graphs fitted to unascertained vs. ascertained datasets. Results are shown for twelve population combinations indicated in plot titles (panels a, b). Results for ascertainment on variants common in Africans (either those having no detectable West Eurasian ancestry or on all Africans in the SGDP dataset) are circled in red. As a starting point for generating different ascertainments, we used either 11,706,773 sites (with no missing data at the group level) polymorphic in a set of 48 archaic and African groups composed of 97 individuals, or 10,051,585 such sites in 59 archaic, African, European, and Middle Eastern groups composed of 120 individuals, or 5,296,653 such sites in 51 Papuan, Native American/Siberian, European, Anatolian, and Caucasian groups composed of 112 individuals (S1 Table). Thirty eight site subsampling schemes were explored (see a list in the legend for Fig 2). The size of the resulting SNP panels is coded by point size, and ten broad ascertainment types are coded by color according to the legends. R2 values for LL are plotted in the left-hand panels, and R2 values for WR are plotted in the right-hand panels. The 2.5th LL or WR percentiles of all the thinned replicates combined, including those on all sites and AT/GC sites, are marked by brown lines. (PDF)