1. I still don’t have a great intuition for why this works–ie. why there is a correlation between LD score and the chi^2 statistic under a polygenic model but not under a stratification model. Specifically, you write that “the more genetic variation an index variant tags, the higher the probability that this index variant will tag a causal variant”. This makes sense, but I feel like I could just as easily write “the more genetic variation an index variant tags, the higher the probability that this index variant will tag a variant that is stratified between the cases and controls for non-causal reasons”. Apparently this isn’t an issue in the simulations, but I wonder if this is generally true, or just true when doing GWAS in closely-related human populations.

2. I’m having a bit of trouble following the math in the supplementary note, though this may be because I’m generally terrible at following math written by other people. For example, the set of steps starting at Eq. 1.3 are totally opaque to me. In the second step there, what are the dimensions of the vectors being multiplied? I feel like something is missing, but again this is probably just me.

Can you show the derivation of the bias in your estimator of r^2 (Eq 1.6)?

]]>Hoping to read in detail some time soon.

]]>