Figure 2

Double cross-validation scheme. In the inner loop, PLS model parameters and variables are estimated based on a bootstrap cross-validation. Based on performance of the PLS models and stability of variables over bootstrap rounds, the optimal parameters and final set of variables are selected. Model performance of the optimized parameters and selected variables are then evaluated on the held-out test set in the outer loop. The outer loop is repeated within a 5-fold cross-validation procedure.