Our F ST computations assume that allele frequencies are obtained from an unrelated set of individuals. If related individuals were used, the effects of sampling error would be underestimated. Unrelated individuals were used in all F ST computations, except in analyses of the aggregate set of Icelandic individuals, which included some related pairs of individuals. In this analysis, we used a subset of 30,244 of the 35,457 Icelandic individuals genotyped, in which the most closely related samples were removed. In this case, the amount by which the estimated sampling error (equal to the reciprocal of N = 2×30,244) is inaccurate is expected to be far smaller than the precision of 0.0001 to which we report F ST estimates, and hence negligible.
