Matrix Math about paper-ldpred2 HOT 4 CLOSED

1. You should not have to match after getting the results from LDpred2, you should always do it before, otherwise you could lose some predictive accuracy.

2. Do not use GFBM = as_FBM(G[,map_pgs2[["_NUM_ID_"]]]), use instead the ind.col parameter in big_prodMat().

3. I am not sure where the issue is but maybe it's because you're multiplying by map_pgs2$beta first, and then only uniformly by (-1)?

Anyway, I'm quite sure there is no issue with big_prodMat(); it must be some mistake in your code.

from paper-ldpred2.

Thanks for the quick reply. I didn't think there was an issue with big_prodMat(). I was trying to understand if it was just performing a simple calculation of the beta value*genotype for each SNP for each individual and then summing them up?

from paper-ldpred2.

Yes, just a matrix product, equivalent to G %*% beta, and yes you need to sum these up if you do it per chromosome.

from paper-ldpred2.

Great. Thank you!

from paper-ldpred2.