SchlossLab/mikropml

Better explain the groups param & batch effects in the intro vignette

Opened this issue · 0 comments

Note from lab meeting code club on the documentation:

There was confusion about how the groups parameter works and why one would want to use it. Need to explicitly state that observations (rows) from the same group are kept together in the train/test split. Overfitting was brought up as a concern -- thinking that using groups might cause models to be overfit. But really it would reveal whether overfitting has occurred.