1: \begin{abstract}

2:   Let $\pi$ denote the intractable posterior density that results when

3:   the likelihood from a multivariate linear regression model with

4:   errors from a scale mixture of normals is combined with the standard

5:   non-informative prior.  There is a simple data augmentation

6:   algorithm (based on latent data from the mixing density) that can be

7:   used to explore $\pi$.  \citet{hobe:jung:khar:2015} recently

8:   performed a convergence rate analysis of the Markov chain underlying

9:   this MCMC algorithm in the special case where the regression model

10:   is \textit{univariate}.  These authors provide simple sufficient

11:   conditions (on the mixing density) for geometric ergodicity of the

12:   Markov chain.  In this note, we extend \pcite{hobe:jung:khar:2015}

13:   result to the multivariate case.

14: \end{abstract}

15: