Is there any way to make/force block.splsda to use a definite group o variables? I mean, I would like to plot, calculate BERs and ERRORs for a given set o variables (class separators). The idea is to compare performance of class separators from different models. Does it make sense?
Thanks for using mixOmics.
To achieve what you have in mind, you can subset your datasets and keep only the variables you are interested in, then run
block.plsda which uses all the variables in all datasets. Afterwards, to assess the performance you can use the
Hope it helps.