I've got two sets of candidate genes and my 'background' of all genes. I think the LD in those two sets differs somewhat from the other, but looking at LD triangle plots doesn't really cut if since there's always a bit of noise there. I can't just take the 'average' LD within those two sets of genes since there are many 0/1 LD SNPs.

regioneR is a nifty package which tests different candidate regions whether they're associated based on overlaps. Is there something like this, but which goes a bit further and checks whether LD is (significantly?) different between two sets of candidate genes? I wouldn't know how to implement that, for starters the SNP density, the SNP missingness, etc. pp. could be different...