Our previous simulations (Table I and supplementary Table I) suggest that application of MDS to IBS metrics calculated from 100,000 uncorrelated SNPs can distinguish fine-scale population structure, which is not apparent when only 10,000 SNPs are used. Supplementary Table II presents the power of the three trend tests of association at a 5% significance level for a high-risk allele frequency of 20% and an allelic odds ratio of 1.5, this time using 100,000 uncorrelated SNPs in calculating the IBS matrix. Comparing these results to those in Table II, we clearly demonstrate a noticeable drop in power to detect association for FST <0.01. With more SNPs, we are better able to distinguish between samples from even relatively closely related populations, with the result that external samples provide little additional information, even at modest levels of population structure. This issue is common to many methods for detecting population structure: a trade-off between sensitivity to fine-scale stratification and correction for its effects in subsequent association analysis.
