Title: Clustering longitudinal ordinal data
Authors: Francesco Amato - University Lyon II (France) [presenting]
Julien Jacques - University Lyon II (France)
Abstract: In social sciences or medicine, studies are often based on questionnaires asking participants to express ordered responses several times over a study period. A model to perform temporal clustering on such longitudinal ordinal data is presented, by assuming that the observed ordinal data are realizations of underlying matrix-normal distributions. Thus, the model relies on a mixture of matrix-variate normal distributions, accounting for the within and between time-dependence structures simultaneously. It allows for possible extension in a mixed data context, to deal jointly with continuous and ordinal data (and possibly more). An EM algorithm for the model estimation is developed. Applications on synthetic and on real data are presented.