Currently bootstrapping can fail when random effects are included in the model. If a subject has very few associated observations, it might be left out the training or test set which results in an error. There should be a ways of resampling that makes sure that every subject appears in every bootstrap dataset. The same problem might occur in cross-validation.
Currently bootstrapping can fail when random effects are included in the model. If a subject has very few associated observations, it might be left out the training or test set which results in an error. There should be a ways of resampling that makes sure that every subject appears in every bootstrap dataset. The same problem might occur in cross-validation.