Title: Model-based profiling of sport preferences
Authors: Rosaria Simone - University of Naples Federico II (Italy) [presenting]
Abstract: Sport preferences and related attitudes can be analysed by collecting ranking or rating data. In these circumstances, it is advisable to adopt a flexible modelling of such discrete distributions that enables the understanding of the evaluation process and the derivation of response profiles. This issue is crucial for marketing purposes, for instance, or when social and behavioural policies have to be addressed as in case sport participation and engagement are investigated. Then, suitable mixture distributions can be specified either to determine clusters of opposite evaluations (for instance, a mixture of inverse hypergeometric models), or to identify structured and guessing response patterns, by choosing a framework in which model parameters are directly linked to explanatory variables. Under this perspective, CUB models lend themselves to advantageous interpretation of results by explicitly accounting for uncertainty. This paradigm can be successfully adapted to incorporate more involved response schemes and to design model-based regression trees and disentangle explanatory features of variables at different subsetting levels. A comparative overview of alternative models based on flexible mixtures of discrete distributions is presented on survey data collected within the BDsport project of the BoDai Lab of University of Brescia, Italy.