1: \begin{abstract}

2: Bayesian analysis of data from the general linear mixed model is

3: challenging because any nontrivial prior leads to an intractable

4: posterior density. However, if a conditionally conjugate prior

5: density is adopted, then there is a simple Gibbs sampler that can be

6: employed to explore the posterior density. A popular default among

7: the conditionally conjugate priors is an improper prior that takes a

8: product form with a flat prior on the regression parameter, and

9: so-called power priors on each of the variance components. In this

10: paper, a convergence rate analysis of the corresponding Gibbs

11: sampler is undertaken. The main result is a simple, easily-checked

12: sufficient condition for geometric ergodicity of the Gibbs--Markov

13: chain. This result is close to the best possible result in the

14: sense that the sufficient condition is only slightly stronger than

15: what is required to ensure posterior propriety. The theory

16: developed in this paper is extremely important from a practical

17: standpoint because it guarantees the existence of central limit

18: theorems that allow for the computation of valid asymptotic standard

19: errors for the estimates computed using the Gibbs sampler.

20: \end{abstract}