Bayesian Inference for Logistic Regression Models Using Sequential Posterior Simulation

The multinomial logistic regression model, hereafter “logit model”, is one of the most widely used models in applied statistics. It provides a straightforward link from covariates to the probabilities of discrete outcomes. More generally, it provides a workable probability distribution for discrete events, whether directly observed or not, as a function of covariates. In the latter, more general, setting it is a key component of conditional mixture models including the mixture of experts models introduced by [23] and studied by [24].