A0259
Title: Mixtures of linear regression models: An application to housing tension in Emilia-Romagna, Italy
Authors: Gabriele Perrone - Department of Statistical Sciences, University of Bologna (Italy)
Gabriele Soffritti - University of Bologna (Italy) [presenting]
Abstract: Mixtures of multivariate linear regression models constitute an approach which simultaneously allows performing linear regression analysis and model-based cluster analysis. They are generally employed when sample observations come from a population composed of unknown sub-populations. Extensions of such models have been recently introduced so as to manage the possible presence of mild outliers and let the researcher be free to use a different vector of covariates for each response. As a consequence, complex real-life data from many areas of activity can be adequately analysed using this type of models. An example is represented by housing deprivation in Italy. As far as the case of the Emilia-Romagna region is concerned, an observatory of the housing system regularly monitors housing conditions and supports the development of public housing policies at a municipality level. Within this framework, a dataset provided by the Emilia-Romagna region has been analysed in order to identify the socio-demographic, income and housing market factors that affect housing tension in the municipalities of the region. The effects of such factors have been studied. Heterogeneity in municipalities has been detected. In order to ensure additional flexibility, the analysis has been complemented by the use of models whose mixing proportions are expressed as functions of some concomitant covariates.
