Need better handling of CV folds if any class contains one sample only for DA analyses

[aljabadi]

We should either prohibit the condition or handle it more efficiently. stratified.subsampling is also unnecessarily lengthy.

data(liver.toxicity)
X <- liver.toxicity$gene
Y <- liver.toxicity$treatment$Dose.Group
# create a class with one sample only
Y[1] <- 'FOO'

liver.plsda <- plsda(X, Y, ncomp = 2)
liver.val <- perf(liver.plsda, validation = "Mfold", folds = 3)
#> Warning in MCVfold.spls(X, Y, multilevel = multilevel, validation = validation, : At least one class is not represented in one fold, which may unbalance the error rate.
#>   Consider a number of folds lower than the minimum in table(Y): 1
#> Error in dimnames(x) <- dn: length of 'dimnames' [2] not equal to array extent

data(nutrimouse)
Y = nutrimouse$diet
# create a class with one sample only
Y = as.character(Y)
Y[1] <- 'FOO'
data = list(gene = nutrimouse$gene, lipid = nutrimouse$lipid)
nutrimouse.sgccda <- block.splsda(X=data,
                                  Y = Y,
                                  keepX = list(gene=c(10,10), lipid=c(15,15)),
                                  ncomp = 2,
                                  scheme = "horst")

perf = perf(nutrimouse.sgccda)
#> Warning in repeat_cv_perf.diablo(nrep): At least one class is not represented in one fold, which may unbalance the error rate.
#>   Consider a number of folds lower than the minimum in table(Y): 1
#> Error in (function (..., deparse.level = 1) : number of columns of matrices must match (see arg 2)
perf
#> function(object, ...)
#>     UseMethod("perf")
#> <environment: namespace:mixOmics>

^{Created on 2021-02-16 by the reprex package (v1.0.0)}