Assuming a prevalence of 1%, equation 5 gives a power of 80% at nominal significance if the explained genetic variance in liability is 7.2%, rising to 99% if the explained genetic variance is 11.7%. Assuming a heritability of 80% [3] this shows that the test was well powered if the marker panel explains about 10% of the heritability, which seems reasonable. The observed result of can be used in equation 6 to give an estimated explained genetic variance of 28.7% (95% CI: 23.6%–33.7%), which is 36% of the heritability, assuming that all SNPs have effects that are identical in the two samples. The estimate reduces only to 26.9% if 99% of the SNPs are assumed null. These results are similar to a recent estimate using mixed modelling of the same data [27].
