Roberto Gismondi

This work remarks the need to carefully evaluate the real importance of each variable used in a multivariate analysis context, with particular regard to cases when an overall performance ranking is the main final purpose. In particular, both a preliminary transformation of variables – aimed at reducing asymmetry and variability of their variation ranges – and the evaluation of their intrinsic explicative power – through redundancy analysis and weighting methods – are proposed. Theoretical and empirical considerations are developed in the frame of quality of life evaluation, carried out at the Italian provinces level on the basis of a yearly survey managed by the Italian economic newspaper ‘‘Il Sole24ore’’. A particular emphasis is given to some normalisation criteria and the case when original variables are grouped ‘‘a priori’’ into logical blocks. A final comparison between the actual ranking method and a series of alternatives is proposed as well.