Do we need X for split

[sachinruk]

Forgive me if this is a dumb question, but if I understand this library correctly, the main aim is to look at correlations between the y's and somehow accomodate for that when stratifying. Is there any reason why we need the X variable as when splitting? eg. to get the indices we always have to do: train_index, test_index = next(iter(msss.split(X, y))).

Thanks in advance.

[trent-b]

For non-stratified splitters, scikit-learn requires X to get the number of samples. For stratified splitters, scikit-learn requires X to get the number of samples for compatibility purposes with the non-stratified splitters although there is no reason I can see that the required y could not be used to get the number of samples instead. For compatibility with scikit-learn, this iterative-stratification package adopts the same practice of requiring X.

[trent-b]

Closing since there has not been additional activity.