fit = function(x, y, wts, param, lev, last, classProbs, ...) {
             ## First fit the pls model, generate the training set scores,
             ## then attach what is needed to the random forest object to
             ## be used later

             ## plsr only has a formula interface so create one data frame
             dat <- x
             dat$y <- y
             pre <- plsr(y~ ., data = dat, ncomp = param$ncomp)
             scores <- predict(pre, x, type = "scores")
             colnames(scores) <- paste("score", 1:param$ncomp, sep = "")
             mod <- randomForest(scores, y, mtry = param$mtry, ...)
             mod$projection <- pre$projection
             mod
             }
           predict = function(modelFit, newdata, submodels = NULL) {
             ## Now apply the same scaling to the new samples
             scores <- as.matrix(newdata)  %*% modelFit$projection
             colnames(scores) <- paste("score", 1:ncol(scores), sep = "")
             scores <- as.data.frame(scores)
             ## Predict the random forest model
             predict(modelFit, scores)
             }