Thank you @danking ! No worries. Would you have any suggestions on how to parallelize logistic_regression_rows() based on its current functionality? With 1500 phenotypes, ~10 seconds/phenotype, we’re looking at 250 minutes, which is not awful.
I’m also not confident I know how to do the error handling properly (so that one bad regression doesn’t halt the entire operation)… I’ll start looking into this too.
Best,
Jeremy