Leave one out?

M7. Evaluating model performance M7. Quiz M7.02
#1

Hi,

I am wondering why ‘leave one out’ wouldn’t be considered in this case? Is it due to bias during the training potentially leading to overfitting?

Thanks,

Merve

#2

LeaveOneOut does not take the group information into account, maybe you thought the question was about LeaveOneGroupOut?

If you don’t take the group information into account, then your estimate of the generalization performance will likely be optimistic. In other words if you have samples of the same hospital both in the train and the test set, your model will likely perform better than if the test set only contain groups that were not seen during training.