Semiparametric multinomial mixed-effects models: A university students profiling tool

Many applicative studies deal with multinomial responses and hierarchical data. Performing clustering at the highest level of grouping, in multilevel regression, is also often interest. In this study we analyse Politecnico di Milano data aim profiling students, modelling their probabilities belonging to different categories considering nested structure within engineering degree programmes. particular, are interested programmes standing on effects types student career. To end, propose an EM algorithm for implementing semiparametric mixed-effects models dealing a response. The novel approach assumes random follow multivariate discrete distribution priori unknown number support points, that is, allowed differ across response categories. advantage twofold: allows, first, express marginal density as weighted sum, avoiding numerical problems integration step, typical parametric approach, and, second, identify latent hierarchy where groups clustered into subpopulations.