EXACT SAMPLING FROM THE STUDENT'S t MODEL

We consider the Bayesian version of the location-scale Student's t model with non-informative prior on the location vector mu and the scale matrix Sigma. Direct analysis of the resulting posterior distribution of (mu, Sigma) is challenging and MCMC methods have been developed in the literature to make draws from this intractable density. In this paper, we propose a method for obtaining an independent and identically distributed (iid) sample from the target posterior density based on the accept-reject method.