Statistical inference with skew t distributions: the mvst r package

We consider a Bayesian analysis of a general regression model with skew-elliptically distributed errors. In particular, we describe the choice of the prior distributions and we propose a Monte Carlo algorithm which allows: i) fast and accurate estimates of the posterior distribution of the parameters ii) to perform model choice among nested families of skew-elliptical classes of densities. All the methods described in the paper are implemented in the new version of the R package mvst. The basic regression model can be modified in several ways. As illustrative examples, we consider a multivariate response model and, with an additional module that interfaces with the package, a stochastic frontier analysis.