Hi there, I have recently run into a problem when dealing with PLSR (and other cross-decomposition methods) prediction output. Unlike other estimators which return a numpy array containing the predictions, PLSR returns a list of single-value lists containing the predictions. I do not know why this done this way (perhaps there is a reason unknown to me) but the fact the estimator is returning a different data type should be advised in the documentation. It is easily solved through modifying the source code or just by overriding the method. Cheers, Jose Carlos Gómez