Approximate verification of geometric ergodicity for multiple-step Metropolis transition kernels

Abstract

In many applications involving discrete time Markov chains, the autocorrelation between states corresponding to nearby time points is too high to use all of these states as part of an approximate random sample from a specified target distribution. In these situations, it is common to use the output of a thinned chain, where we take samples every $h$ steps, and $h$ is a positive integer, in order to reduce autocorrelation. In order to justify using central limit theorems in analyses based on the output of a thinned chain, it is necessary to show that the thinned chain is geometrically ergodic. A common way to do this is to show that the chain satisfies a minorization condition and an associated drift condition. In this manuscript, we extend previous results pertaining to one-step transition kernels to handle numerical estimation of minorization and drift coefficients for $h$-step transition kernels for Metropolis algorithms.

Keywords

• Markov chain Monte Carlo
• Geometric Ergodicity
• drift
• minorization

References

1. [1] Y.F. Atchadé, Approximate spectral gaps for Markov chain mixing times in high dimensions, SIMODS 3 (3), 854-872, 2021.
2. [2] S. Chib, F. Nardari and N. Shephard, Markov chain Monte Carlo methods for generalized stochastic volatility models, J. Econometrics 108 (2), 281–316, 2002.
3. [3] M.K. Cowles and J.S. Rosenthal, A simulation-based approach to convergence rates for Markov chain Monte Carlo algorithms, Statist. Comput. 8, 115–124, 1998.
4. [4] G. Fort, E. Moulines, G.O. Roberts and J.S. Rosenthal, On the geometric ergodicity of hybrid samplers, J. Appl. Probab. 40 (1), 123–146, 2003.
5. [5] A. Gelman, W.R. Gilks and G.O. Roberts, Weak convergence and optimal scaling of random walk Metropolis algorithms, Ann. Appl. Probab. 7 (1), 110–120, 1997.
6. [6] A. Gelman and D.B. Rubin, Inference from iterative simulation using multiple sequences, Statist. Sci. 7 (4), 457–511, 1992.
7. [7] J. Geweke, Evaluating the accuracy of sampling-based approaches to calculating posterior moments, in: Bayesian Statistics 4, Eds: J.M. Bernardo, J. Berger, A.P. Dawid and A.F.M. Smith, Oxford University Press, 169–193, 1992.
8. [8] P. Heidelberger and P.D. Welch, Simulation run length control in the presence of an initial transient, Oper. Res. 31 (6), 1109–1144, 1983.

9. [9] H. Ishwaran, L.F. James and J. Sun, Bayesian model selection in finite mixtures by marginal density decompositions, J. Amer. Statist. Assoc. 96 (456), 1316–1332, 2001.
10. [10] S.F. Jarner and E. Hansen, Geometric ergodicity of Metropolis algorithms, Stochastic Process. Appl. 85 (2), 341–361, 2000.
11. [11] V.E. Johnson, Studying convergence of Markov chain Monte Carlo algorithms using coupled sampling paths, J. Amer. Statist. Assoc. 91 (433), 154–166, 1996.
12. [12] F. Liang, Continuous contour Monte Carlo for marginal density estimation with an application to a spatial statistical model, J. Comput. Graph. Statist. 16 (3), 608–632, 2007.
13. [13] S.P. Meyn and R.L. Tweedie Markov Chains and Stochastic Stability, 2nd ed., Springer-Verlag, London, 2005.
14. [14] R.M. Neal, Annealed Importance Sampling, Technical report, University of Toronto, Department of Statistics, 1998.
15. [15] M. Oh and J.O. Berger, Adaptive importance sampling in Monte Carlo integration, Technical report, Purdue University, Department of Statistics, 1989.
16. [16] G.O. Roberts, Methods for estimating L2 convergence of Markov chain Monte Carlo, in: Bayesian Statistics and Econometrics: Essays in Honor of Arnold Zellner, Eds: D. Berry, I. Chaloner and J. Geweke, Amsterdam, North-Holland, 373–384, 1996.
17. [17] G.O. Roberts and J.S. Rosenthal, Coupling and ergodicity of adaptive Markov chain Monte Carlo algorithms, J. Appl. Probab. 44 (2), 458-475, 2007.
18. [18] G.O. Roberts and R.L. Tweedie, Geometric convergence and central limit theorems for multidimensional Hastings and Metropolis algorithms, Biometrika 83 (1), 95– 110, 1996.
19. [19] J.S. Rosenthal, Minorization conditions and convergence rates for Markov chain Monte Carlo, J. Amer. Statist. Assoc. 90 (430), 558–566, 1995.
20. [20] D.A. Spade, A computational procedure for efficient estimation of the mixing time of a random-scan Metropolis algorithm, Stat. Comput. 26 (4), 761–781, 2016.
21. [21] D.A. Spade, A computational approach to bounding the mixing time of a Metropolis- Hastings sampler, Markov Process. Relat. Fields 26 (3), 487–516, 2020.
22. [22] D.A. Spade, A Monte Carlo integration approach to estimating drift and minorization coefficients for Metropolis-Hastings samplers, Braz. J. Probab. Stat. 35 (3), 466–483, 2021.
23. [23] B. Yu, Monitoring the convergence of Markov samplers based on estimated L1 error, Technical report 409, University of California, Department of Statistics, 1994.
24. [24] B. Yu and P. Mykland, Looking at Markov samplers through CUSUM path plots: a simple diagnostic idea, Technical report 413, University of California, Department of Statistics, 1994.