Andrea Cremaschi, professor at IE University Madrid, has been the featured speaker at the Statistics Seminar Series at the University of Bergamo. His talk, titled “Matrix-Variate Priors for Flexible Mixture Modelling of Grouped Data” explored innovative approaches to Bayesian nonparametric modeling, offering new perspectives on dependent prior distributions for mixture models.

In collaboration with B. Franzolini, Professor Cremaschi introduced a finite-dimensional matrix-variate Dirichlet distribution as a novel framework for Bayesian mixture models. This method extends beyond standard univariate priors by allowing for greater flexibility, structured dependency across groups, and enhanced interpretability. Unlike conventional approaches that primarily rely on hierarchical nonparametric processes, this model facilitates group-specific kernel selection while ensuring robust inference through a tailored MCMC algorithm. The proposed methodology has broad applications, with demonstrated effectiveness in both simulated and real-world datasets.