Contributions of Italian statisticians to the development of multivariate data analysis

The main contributions of Italian statisticians to the methodology of multivariate data analysis are investigated, focusing specifically on the development of techniques for coping with the extraction of information from complex data characterized by two or more variables or sets of variables as observed on one or more sets of objects. In particular the following types of methodological areas are considered: supervised and unsupervised classification, regression, factorial and scaling approaches. Methods for dealing with different sources of uncertainty associated with the procedures for drawing information from the data are examined with reference to: sampling uncertainty, model uncertainty, imprecision/vagueness of the data. The Italian contributions are discussed in the framework of various lines of research, including: analysis of contingency tables, asymmetric relationships among sets of variables, multiway data analysis, fuzzy and symbolic data analysis, textual analysis, stability analysis and model selection. Although the bulk of this study is devoted to the works appeared in the last three or four decades, some hints are given to the historical profile of the Italian school of Statistics. In this connection it is underlined that the more recent developments are characterized by specific traits of originality, which place the Italian contributions to the aforementioned fields of research somehow at the crossroads among the French, the Dutch and the Anglo-American schools of Statistics.