In this presentation, we propose a decision-theoretic approach that brings to light the relation between a sparse loadings matrix and the factor dimension. This relation is done through a summary from the information contained in the multivariate posterior. Under the presented approach, the prior is a component of the Bayesian specification while model selection is restricted to a loss function that reflects the trade-off between fit and sparsity. This trade-off is then displayed in a summary that exposes the models with the best fit. We compare this new procedure with other recent approaches through a simulation study and application to real data.
Após o registro, você receberá um e-mail de confirmação contendo informações sobre conexão no seminário.
Henrique Bolfarine - holds a PhD in Statistics from the University of São Paulo, a BA in Applied Mathematics, and a MA in Statistics from the University of São Paulo. Currently he is a researcher at the Salem Center for Policy, at the University of Texas at Austin. His research interests include Bayesian variable selection, MCMC methods, mixture models, factor models, Bayesian learning, and applications in social sciences.