The Generalized Dynamic Mixtures of Factor Analyzers for clustering multivariate longitudinal data

This work introduces the Generalized Dynamic Mixtures of Factor Analyzers (GDMFA) approach for clustering high-dimensional longitudinal data. The proposed model can be seen as an extension of the Gaussian mixture model where individuals are allowed to move between components over time and, within each component, local dimensional reduction is performed. Temporal dependence is modelled through a first-order finite-state Markov chain. The model parameters have been estimated through an Alternating Expected Conditional Maximization (AECM) algorithm and the performance of the GDMFA model is discussed on the equitable and sustainable well-being (BES) of Italian territories data set. The results are encouraging and would deserve further discussion.