Dividing dataset to create model

Dear mixOmics developer,
I create a model using your R-package, I use PLS-DA function. This function provide m-folds cross validation. In this case, when we have a small sample dataset, does it need to be divided into training and test set? Considering that in the function it is cross-validated already.

F.S. Aurum

hi @aurum,

If your sample size is very small and you are not interested in assessing the performance of PLS-DA or tune the model, then you can skip the functions perf() or tune().

If not, then you can use leave-one-out cross validation by changing the argument ‘validation’ in these functions (cross-validation is only implemented in these two functions).


Thank you Prof Kim-Anh