A collinearity based hierarchical method to identify clusters of variables

Abstract

The most frequently used hierarchical methods for clustering of quantitative variables are based on bivariate or multivariate correlation measures. These solutions can be unsuitable in presence of uncorrelated but collinear variables. In this paper we propose a hierarchical agglomerative algorithm based on a similarity measure which takes into account the collinearity between two groups of variables. Its main theoretical features are described and its performance is evaluated both on simulated and real data sets.