A0367
Title: Mixture-based clustering for ordinal responses
Authors: Marta Nai Ruscone - Università degli Studi di Genova (Italy) [presenting]
Abstract: Existing methods can perform likelihood-based clustering on a multivariate data matrix of ordinal data, using finite mixtures to cluster the rows (observations) of the matrix. These models can incorporate the main effects of individual rows and columns, as well as cluster effects, to model the matrix of responses. However, many real-world applications also include available covariates, which can provide insights into the main characteristics of the clusters. The mixture-based models are extended to include covariates directly, to allow the clustering structures to be determined both by the individuals' similar patterns of responses and the effects of the covariates on the individuals' responses. The focus is on clustering the rows of the data matrix, using the proportional odds cumulative logit model for ordinal data. The models are fit using the expectation-maximization algorithm, and performance is assessed through a comprehensive simulation study. An application of the models is also illustrated.
