Cluvar: a grouping technique of statistic units and of variables selection

Problems related to the identification of the complex of variables are present in researches regarding the division -related to many phenomena- of a territory in homogeneous areas. The complex of variables must individuate the hypothesis in a univocal way and define the relation among the areas. On the other hand the complex of definite indicators and the kind of grouping methods hypothesised strongly conditioned the choice of a technique of classification. The most common techniques of statistical classification need the previous determination of the groups' number or of the elements number per every group. The goal of this research is to define a technique of classification that allows the formation of homogeneous groups without the previous definition of their numerousness. At the same time this classification make possible the selection of the variables that determine the most significant links among the unities. This technique gives the description of the relations among the single units and the groups. In this research this methodology is used to determine the regional diversification of the Italian financial system. With the cluster analysis applied to the territorial unities we obtain the structural elements which allow the individuation of indicators' groups of real and financial economy that synthesize both the information obtained from the index and a system of relation among the considered territorial units.