Processing math: 100%
Skip to content
Licensed Unlicensed Requires Authentication Published by De Gruyter August 8, 2021

Estimating drift and minorization coefficients for Gibbs sampling algorithms

  • David A. Spade EMAIL logo

Abstract

Gibbs samplers are common Markov chain Monte Carlo (MCMC) algorithms that are used to sample from intractable probability distributions when sampling directly from full conditional distributions is possible. These types of MCMC algorithms come up frequently in many applications, and because of their popularity it is important to have a sense of how long it takes for the Gibbs sampler to become close to its stationary distribution. To this end, it is common to rely on the values of drift and minorization coefficients to bound the mixing time of the Gibbs sampler. This manuscript provides a computational method for estimating these coefficients. Herein, we detail the several advantages of the proposed methods, as well as the limitations of this approach. These limitations are primarily related to the “curse of dimensionality”, which for these methods is caused by necessary increases in the numbers of initial states from which chains need be run and the need for an exponentially increasing number of grid points for estimation of minorization coefficients.

MSC 2010: 60J05

References

[1] M. K. Cowles and J. S. Rosenthal, A simulation-based approach to convergence rates for Markov Chain Monte Carlo algorithms, Statist. Comput. 8 (1998), 115–124. 10.1023/A:1008982016666Search in Google Scholar

[2] T. Erkkilä, S. Lehmusvaara, P. Ruusuvuori, T. Visakorpi, I. Shmulevich and H. Lähdesmäki, Probabilistic analysis of gene expression from heterogeneous tissues, Bioinform. 26 (2010), no. 20, 2571–2577. 10.1093/bioinformatics/btq406Search in Google Scholar

[3] G. Fort, E. Moulines, G. O. Roberts and J. S. Rosenthal, On the geometric ergodicity of hybrid samplers, J. Appl. Probab. 40 (2003), no. 1, 123–146. 10.1239/jap/1044476831Search in Google Scholar

[4] A. Gelman and D. B. Rubin, Inference from iterative simulation using multiple sequences, Statist. Sci. 7 (1992), 457–511. 10.1214/ss/1177011136Search in Google Scholar

[5] J. Geweke, Evaluating the accuracy of sampling-based approaches to the calculation of posterior moments, Bayesian Statistics 4 (Peñíscola 1991), Oxford University, New York (1992), 169–193. 10.21034/sr.148Search in Google Scholar

[6] R. J. B. Goudie and S. Mukherjee, A Gibbs sampler for learning DAGs, J. Mach. Learn. Res. 17 (2016), Paper No. 30. Search in Google Scholar

[7] P. Heidelberger and P. D. Welch, Simulation run length control in the presence of an initial transient, Oper. Res. 31 (1983), 1109–1144. 10.1287/opre.31.6.1109Search in Google Scholar

[8] T. Hrycej, Gibbs sampling in Bayesian networks, Artificial Intelligence 46 (1990), no. 3, 351–363. 10.1016/0004-3702(90)90020-ZSearch in Google Scholar

[9] S. R. F. Jarner and E. Hansen, Geometric ergodicity of Metropolis algorithms, Stochastic Process. Appl. 85 (2000), no. 2, 341–361. 10.1016/S0304-4149(99)00082-4Search in Google Scholar

[10] Y. Liu, O. Simeone, A. M. Haimovich and W. Su, Modulation classification via Gibbs sampling based on a latent Dirichlet Bayesian network, IEEE Sig. Proc. Lett. 21 (2014), no. 9, 1135–1139. 10.1109/LSP.2014.2327193Search in Google Scholar

[11] K. L. Mengersen and R. L. Tweedie, Rates of convergence of the Hastings and Metropolis algorithms, Ann. Statist. 24 (1996), no. 1, 101–121. 10.1214/aos/1033066201Search in Google Scholar

[12] S. P. Meyn and R. L. Tweedie, Markov Chains and Stochastic Stability, 2nd ed., Springer, London, 2005. Search in Google Scholar

[13] A. E. Raftery and S. Lewis, How many iterations in the Gibbs sampler?, Bayesian Statistics 4, Oxford University, Oxford (1992), 763–773. 10.21236/ADA640705Search in Google Scholar

[14] G. O. Roberts, Convergence diagnostics of the Gibbs sampler, Bayesian Statistics 4 (Peñíscola 1991), Oxford University, New York (1992), 775–782. 10.1093/oso/9780198522669.003.0054Search in Google Scholar

[15] G. O. Roberts, Methods for estimating L2 convergence of Markov Chain Monte Carlo, Bayesian Statistics and Econometrics: Essays in Honor of Arnold Zellner, North-Holland, Amsterdam (1996), 373–384. Search in Google Scholar

[16] G. O. Roberts and J. S. Rosenthal, Geometric ergodicity and hybrid Markov chains, Electron. Commun. Probab. 2 (1997), Paper No. 2. 10.1214/ECP.v2-981Search in Google Scholar

[17] G. O. Roberts and J. S. Rosenthal, Two convergence properties of hybrid samplers, Ann. Appl. Probab. 8 (1998), no. 2, 397–407. 10.1214/aoap/1028903533Search in Google Scholar

[18] G. O. Roberts and J. S. Rosenthal, Convergence of slice sampler Markov chains, J. R. Stat. Soc. Ser. B Stat. Methodol. 61 (1999), no. 3, 643–660. 10.1111/1467-9868.00198Search in Google Scholar

[19] J. S. Rosenthal, Minorization conditions and convergence rates for Markov chain Monte Carlo, J. Amer. Statist. Assoc. 90 (1995), no. 430, 558–566. 10.1080/01621459.1995.10476548Search in Google Scholar

[20] M. D. Sonksen, X. Wang and K. Umland, Bayesian partially ordered multinomial probit and logit models with an application to course redesign, Technical Report, University of New Mexico, 2013. Search in Google Scholar

[21] D. A. Spade, A computational procedure for estimation of the mixing time of the random-scan Metropolis algorithm, Stat. Comput. 26 (2016), no. 4, 761–781. 10.1007/s11222-015-9568-3Search in Google Scholar

[22] D. A. Spade, A computational approach to bounding the mixing time of a Metropolis–Hastings sampler, Markov Process. Related Fields 26 (2020), no. 3, 487–495. Search in Google Scholar

[23] D. A. Spade, Geometric ergodicity of a Metropolis–Hastings algorithm for Bayesian inference of phylogenetic branch lengths, Comput. Statist. 35 (2020), no. 4, 2043–2076. 10.1007/s00180-020-00969-1Search in Google Scholar

[24] D. A. Spade, A Monte Carlo integration approach to estimating drift and minorization coefficients for Metropolis–Hastings samplers, Braz. J. Probab. Stat. 35 (2021), no. 3, 466–483. 10.1214/20-BJPS486Search in Google Scholar

[25] D. A. Spade, R. Herbei and L. S. Kubatko, Geometric ergodicity of a hybrid sampler for Bayesian inference of phylogenetic branch lengths, Math. Biosci. 268 (2015), 9–21. 10.1016/j.mbs.2015.07.002Search in Google Scholar PubMed

[26] B. Yu, Monitoring the convergence of Markov samplers based on estimated L1 error, Technical Report 409, University of California, 1994. Search in Google Scholar

[27] B. Yu, P. Mykland, Looking at Markov samplers through CUSUM path plots: A simple diagnostic idea, Technical Report 413, University of California, 1994. Search in Google Scholar

[28] A. Zellner and C. K. Min, Gibbs sampler convergence criteria, J. Amer. Statist. Assoc. 90 (1995), 921–927. 10.1080/01621459.1995.10476591Search in Google Scholar

Received: 2021-01-04
Revised: 2021-07-15
Accepted: 2021-07-21
Published Online: 2021-08-08
Published in Print: 2021-09-01

© 2021 Walter de Gruyter GmbH, Berlin/Boston

Downloaded on 18.2.2025 from https://www.degruyter.com/document/doi/10.1515/mcma-2021-2093/html
Scroll to top button