I am trying to do genomic control on the simulated data generated from the Balding-Nichols model (Fst = 0.01) on two populations. The lamba which I am obtaining after dividing the median chi sq statistic on all of the SNPs by median of chisq(1) is quite big.
The p-value threshold I've set is 1e-4 and the chi sq statistic for this p-value and 1 degree of freedom is 15.13. However, in all scenarios due to the lambda being quite big (around 60 and the max uncorrected chi sq statistic value is around 70), all of the corrected chi sq statistic is far below the p-value threshold. Am I missing something?
It'd be great if anybody can help me understand this better.