Different views of the multivariate ranking problem

Abstract

Multivariate ranking problems are characterized by the need of ordering C different items according to several different features. The multivariate nature of these problems makes them quite challenging and flexible multivariate statistical techniques are therefore required.

In this study we focus on two different scenarios, where we need to rank C different populations. Under the first scenario, preliminary knowledge about the order of the populations is available, while under the second one no information is available.

Two solutions, based on the Nonparametric combination (NPC) technique, are proposed to deal with these scenarios and two case studies are adopted to facilitate the comprehension of the methods and to highlights the main differences between the two considered multivariate ranking problems.