Hi there,
I am working with some clinical data as well as MRI data. I would like to know if I can use sPLS method in Mixomics package for binary or categorical input data. In fact, I know that I could consider categorical outcomes, but I am really not sure if it also accepts categorical or binary variables in the x matrix.
Hello AFJAM,
We confirm that binary or categorical input data are going to pose a problem as they will be considered as continuous in sPLS-DA. You could transform each categorical variable as a dummy matrix (by using the unmap() function in mixOmics) but if you use a sparse PLS-DA the interpretation is going to be tricky (as it may select one category of a given variable, but not the other). Worth a try.
Otherwise there are other non parametric methods (e.g. random forest) that might be more suitable.
Kim-Anh
PS: you could have a read at tip 3 in this paper: https://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1006907
I was proposing an MCA type approach wit the dummy matrices, but multifactor analysis, or other approaches are also possible (outside mixOmics).
I gave an example in this post here on how to transform dummy matrices