I've been investigating ways of looking for epistasis in GWAS data, with the intention of looking for interactions with a specific SNP (hence the --condition option used above). From my tests so far, the above options appear to replicate results fairly closely to methods I can use in R, and also the --epistasis module in PLINK, while still allowing me to apply additional covariates (not allowed with --epistasis) and to easily carry out on a genome wide scale (my approach in R was on a per-interaction basis).
My question two-fold. Firstly, is this combination of commands appropriate as a replacement for --epistasis? I'm effectively just doing the same kind of logistic regression after all from what I can see.
Secondly, in order to get anything other than NA's throughout the results file, I had to raise the VIF using --vif, as recommended in the PLINK manual here. However simply raising it to a VIF of 100 didn't help, trying it at --vif 1000 finally gave me output (though towards the end of the file I still got many NAs). How does applying the VIF modification affect my results? Are the results still valid? Will raising the VIF even further get rid of remaining NAs? Would that even be necessary? (As the top results I get from what I have so far seem to make sense in context)
