There are two variable importance measures in PLS-DA:
-
Variable Importance in Projection (VIP), which is a weighted sum of squares of the PLS loadings taking into account the amount of explained Y-variation, in each dimension (i.e. VIP scores are calculated for each component).
-
Weighted sum of absolute regression coefficients. The weights are a function of the reduction of the sums of squares across the number of PLS components. Please note, for multiple-group (more than two) analysis, the same number of predictors will be built for each group. Therefore, the coefficient of each feature will be different depending on which group you want to predict. The average of the feature coefficients are used as the overall coefficient-based importance.