search

bvilhjal / ldpred

MIT License
95 stars 58 forks source link

prs adjustments #79

Closed bnj50 closed 5 years ago

bnj50 commented 5 years ago

In the output score files, when I use the last column of score files to generate auc (called as : cov-sex-pc-adj), I am getting much better AUC-performance than when I just use PRS only… that sounds like the last column represent the total variation not “the PRS adjusted for cov-sex-pc” . In that case the last term (adj) is not correct in output...it should be called "total variations"

IID true_phens PRS cov_adj pc_adj cov_sex_pc_adj

Please let us know about this … thanks

bvilhjal commented 5 years ago

Thanks for your comment, and I apologize for the slow reply (I was on holidays in April).

You're right, these labels are wrong (or at least misleading). These are not adjusted prediction, rather total predictions. I updated the labels to reflect this.

More generally, I currently recommend just importing the PRS into some other software, e.g. R, where you can study the relative importance of PCs, sex, and other covariates with and without the PRS.

Best, Bjarni