As expected, the false-positive error rate of T_CC is correctly maintained at the significance level, regardless of FST, since these samples are ascertained from the same population. Expanding the controls with external samples dramatically inflates the false-positive error rate of T_F, as the external cohorts become increasingly divergent from the case-control source population. The test becomes biased in an anticonservative direction for FST, even as low as 0.002, which we might expect between populations of Northern and Western European ancestry, for example. For T_CC and T_Fmds, the mean allelic odds ratios are unbiased estimates of the heterozygous genotype relative risk, as expected under a multiplicative disease model, whilst for T_F there is some inflation for the mean extreme models of population structure.
