Thanks @Neystale
I confirm what @Neystale said (thanks!). PLS-DA is not great with ordinal data. For an ordinal response, see this discussion here: .
For ordinal predictors: Is data suitable for mixomics? - #2 by kimanh.lecao. You could look at all of them at the same time, but they would be considered as continuous.
None of the options discussed above are great workaround though!
Kim-Anh