Robust multivariate methods for the analysis of the university performance

One of the most important problems among the methodological issues discussed in cluster analysis is the identification of the correct number of clusters and the correct allocation of units to their natural clusters. In this paper we use the forward search algorithm, recently proposed by Atkinson, Riani and Cerioli (2004) to scrutinize in a robust and efficient way the output of k-means clustering algorithm. The method is applied to a data set containing efficiency and effectiveness indicators, collected by the National University Evaluation Committee (NUEC), used to evaluate the performance of Italian universities.