Then, the following methods were performed on all the re-duced matrices obtained:
a) Principal Component Analysis (PCA) as a display method, in order to visualize the data structure.
b) LDA as a classification technique, in order to classify the oil samples on the basis of their geographical origin.
c) QDA-UNEQ as a class-modelling technique, in order to build models for Chianti Classico oil.
Then, the fusion of information coming from the three instru-ments was performed as a strategy to increase the classification and class-modelling results. In order to reach this goal, all the variables selected by STEP-LDA from the 14 matrices were joined and STEP-LDA was applied again on this subset of fused variables to retain the most discriminant ones.