Closed
Description
As discussed per mail, it would be great, if kfold in brms allowed something similar to the folds argument in rstanarm, which allows us to ensure that dependent data points end up in the same fold (when that is desired).
A simple way to implement this with the present arguments would be to allow the simultaneous specification of both K and group. When a group variable is given, K stratified folds would be formed based on that group. Currently, K is ignored, when specifying group.
For other, more complex ways of defining folds (e.g., for timeseries), it might be useful to allow for a more general user-defined folds argument as in rstanarm that simply allows us to assign rows to a fold.