The Number of MCMC Draws Needed to Compute Bayesian Credible Bounds

ABSTRACT In the past 20 years, there has been a staggering increase in the use of Bayesian statistical inference, based on Markov chain Monte Carlo (MCMC) methods, to estimate model parameters and other quantities of interest. This trend exists in virtually all areas of engineering and science. In a typical application, researchers will report estimates of parametric functions (e.g., quantiles, probabilities, or predictions of future outcomes) and corresponding intervals from MCMC methods. One difficulty with the use of inferential methods based on Monte Carlo (MC) is that reported results may be inaccurate due to MC error. MC error, however, can be made arbitrarily small by increasing the number of MC draws. Most users of MCMC methods seem to use indirect diagnostics, trial-and-error, or guess-work to decide how long to run a MCMC algorithm and accuracy of MCMC output results is rarely reported. Unless careful analysis is done, reported numerical results may contain digits that are completely meaningless. In this article, we describe an algorithm to provide direct guidance on the number of MCMC draws needed to achieve a desired amount of precision (i.e., a specified number of accurate significant digits) for Bayesian credible interval endpoints.

[1]  R. Woodruff,et al.  Confidence Intervals for Medians and Other Position Measures , 1952 .

[2]  N. Metropolis,et al.  Equation of State Calculations by Fast Computing Machines , 1953, Resonance.

[3]  W. K. Hastings,et al.  Monte Carlo Sampling Methods Using Markov Chains and Their Applications , 1970 .

[4]  E. Nummelin,et al.  A splitting technique for Harris recurrent Markov chains , 1978 .

[5]  K. Athreya,et al.  A New Approach to the Limit Theory of Recurrent Markov Chains , 1978 .

[6]  Ing Rj Ser.  Approximation Theorems of Mathematical Statistics , 1980 .

[7]  Bruce W. Schmeiser,et al.  Batch Size Effects in the Analysis of Simulation Output , 1982, Oper. Res..

[8]  Donald Geman,et al.  Stochastic Relaxation, Gibbs Distributions, and the Bayesian Restoration of Images , 1984, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[9]  M. R. Adams,et al.  Measure Theory and Probability , 1986 .

[10]  A. Dobson,et al.  Epidemiology of acquired immunodeficiency syndrome in Australia , 1987, The Medical journal of Australia.

[11]  Adrian F. M. Smith,et al.  Bayesian computation via the gibbs sampler and related markov chain monte carlo methods (with discus , 1993 .

[12]  H. Künsch.  The Jackknife and the Bootstrap for General Stationary Observations , 1989 .

[13]  G. Rutherford,et al.  The epidemiology of acquired immunodeficiency syndrome. , 1989, Immunology series.

[14]  Adrian F. M. Smith,et al.  Sampling-Based Approaches to Calculating Marginal Densities , 1990 .

[15]  A. Raftery,et al.  How Many Iterations in the Gibbs Sampler , 1991 .

[16]  D. Andrews.  Heteroskedasticity and Autocorrelation Consistent Covariance Matrix Estimation , 1991 .

[17]  Charles J. Geyer,et al.  Practical Markov Chain Monte Carlo , 1992 .

[18]  Joseph P. Romano,et al.  On the sample variance of linear statistics derived from mixing sequences , 1993 .

[19]  Peter Hall,et al.  Smoothed empirical likelihood confidence intervals for quantiles , 1993 .

[20]  L. Tierney.  Markov Chains for Exploring Posterior Distributions , 1994 .

[21]  Joseph P. Romano,et al.  BIAS‐CORRECTED NONPARAMETRIC SPECTRAL ESTIMATION , 1995 .

[22]  Bin Yu,et al.  Regeneration in Markov chain samplers , 1995 .

[23]  Sylvia Richardson,et al.  Markov chain concepts related to sampling algorithms , 1995 .

[24]  S. Chib,et al.  Understanding the Metropolis-Hastings Algorithm , 1995 .

[25]  Theodore P. Hill,et al.  Strongly-Consistent, Distribution-Free Confidence Intervals for Quantiles , 1997 .

[26]  S. Lahiri.  Theoretical comparisons of block bootstrap methods , 1999 .

[27]  M. Kosorok.  Monte Carlo error estimation for multivariate Markov chains , 2000 .

[28]  Randy R. Sitter,et al.  A note on Woodruff confidence intervals for quantiles , 2001 .

[29]  Eric R. Ziegel,et al.  An Introduction to Generalized Linear Models , 2002, Technometrics.

[30]  Galin L. Jones,et al.  On the applicability of regenerative simulation in Markov chain Monte Carlo , 2002 .

[31]  S. Lahiri.  Resampling Methods for Dependent Data , 2003 .

[32]  D. Politis.  Adaptive bandwidth choice , 2003 .

[33]  Martyn Plummer,et al.  JAGS: A program for analysis of Bayesian graphical models using Gibbs sampling , 2003 .

[34]  Galin L. Jones.  On the Markov chain central limit theorem , 2004, math/0409112.

[35]  Galin L. Jones,et al.  Fixed-Width Output Analysis for Markov Chain Monte Carlo , 2006, math/0601446.

[36]  K. Athreya,et al.  Measure Theory and Probability Theory , 2006 .

[37]  P. Bertail,et al.  Regenerative block bootstrap for Markov chains , 2006 .

[38]  Murali Haran,et al.  Markov chain Monte Carlo: Can we trust the third significant figure? , 2007, math/0703746.

[39]  David J. Earl,et al.  Monte Carlo simulations. , 2008, Methods in molecular biology.

[40]  Peter D. Hoff,et al.  A First Course in Bayesian Statistical Methods , 2009 .

[41]  S. Haneuse,et al.  On the Assessment of Monte Carlo Error in Simulation-Based Statistical Analyses , 2009, The American statistician.

[42]  Fang Chen.  Bayesian Modeling Using the MCMC Procedure , 2009 .

[43]  James M. Flegal,et al.  Batch means and spectral variance estimators in Markov chain Monte Carlo , 2008, 0811.1729.

[44]  Jesse Frey,et al.  Fixed-Width Sequential Confidence Intervals for a Proportion , 2010 .

[45]  H. Harari-Kermadec.  Regenerative block empirical likelihood for Markov chains , 2011, 1102.3107.

[46]  Andrew Gelman,et al.  Inference from Simulations and Monitoring Convergence , 2011 .

[47]  James M. Flegal,et al.  Markov chain Monte Carlo estimation of quantiles , 2012, 1207.6432.

[48]  James M. Flegal,et al.  Relative fixed-width stopping rules for Markov chain Monte Carlo simulations , 2013, 1303.0238.

[49]  Hani Doss,et al.  Estimates and standard errors for ratios of normalizing constants from multiple Markov chains via regeneration , 2014, Journal of the Royal Statistical Society. Series B, Statistical methodology.

[50]  Bernd Weissmuller.  Measure Theory And Probability , 2016 .

[51]  ′ Mixing , .