Open dereksonderegger opened 6 years ago
I think the best thing to do is to use glmnet() somewhat directly. In particular, I think we ought to run it through the package glmnetUtils
which gives us a formula based interface and a predict() function that isn't insane.
There was a great deal of students that couldn't get this to work remotely consistently for picking best lambda and the 1 SE. Passing in lambda twice is an issue to fix. There is a function OneSE that I couldn't quite get right.