Hi All, I am using the cppls function in the pls package, and I want to use cross validation to determine the best number of components. Since Hastie et al recommended a "one standard error rule", i.e., choose the most parsimonious model whose error is no more than one standard error above the error of the best model, I am wondering how I can get the standard error of misclassification rate from pls package? Thank you, Cindy [[alternative HTML version deleted]]