How Many Iterations in the Gibbs Sampler?

Abstract

When the Gibbs sampler is used to estimate posterior distributions (Gelfand and Smith, 1990), the question of how many iterations are required is central to its implementation. When interest focuses on quantiles of functionals of the posterior distribution, we describe an easily-implemented method for determining the total number of iterations required, and also the number of initial iterations that should be discarded to allow for \burn-in". The method uses only the Gibbs iterates themselves, and does not, for example, require external speci cation of characteristics of the posterior density. Here the method is described for the situation where one long run is generated, but it can also be easily applied if there are several runs from di erent starting points. It also applies more generally to Markov chain Monte Carlo schemes other than the Gibbs sampler. It can also be used when several quantiles are to be estimated, when the quantities of interest are probabilities rather than full posterior distributions, and when the draws from the posterior distribution are required to be approximately independent. The method is applied to several di erent posterior distributions. These include a multivariate normal posterior distribution with independent parameters, a bimodal distribution, a \cigar-shaped" multivariate normal distribution in ten dimensions, and a highly complex 190-dimensional posterior distribution arising in spatial statistics. In each case the method appears to give satisfactory results. The results suggest that reasonable accuracy may often be achieved with 5,000 iterations or less; this can frequently be reduced to less than 1,000 if the posterior tails are known to be light. However, there are frequent \exceptions" when the required number of iterations is much higher. One important such exception is when there are high posterior correlations between the parameters; even crude correlation-removing This research was supported by ONR contract N-00014-88-K-0265 and by NIH grant no. 5R01HD26330-02. The authors are grateful to Jeremy York for providing the data for Examples 4 and 5, for helping with the analysis and for useful discussions and suggestions, and to Julian Besag and an anonymous referee for helpful comments. A Fortran program called \Gibbsit" that implements the methods described here may be obtained from StatLib by sending an e-mail message to

Genetic variation and effects of candidate-gene polymorphisms on coagulation properties, curd firmness modeling and acidity in milk from Brown Swiss cows

“…The marginal posterior distributions of all parameters were obtained using a Gibbs sampler running with a single chain of 1 000 000 points; the first 50 000 were discarded as burnin, as previously tested by Raftery and Lewis (1992). Samples were saved every 100 iterations.…”

Comparison between linear and proportional hazard models for the analysis of age at first lambing in the Ripollesa breed

“…Three independent Monte Carlo Markov chains (MCMC) with 1 000 000 iterations were run for each analysis, and the first 100 000 iterations were discarded as burn-in. The suitability of the length of the burn-in period was evaluated by both visual inspection and the method of Raftery and Lewis (1992) on the sampling path of σ a 2 . A total of 18 000 samples of model parameters were saved from each chain with a lag interval of 50 iterations.…”

Sources of sire-specific genetic variance for birth and weaning weight in Bruna dels Pirineus beef calves

“…After exploratory analyses, we launched three independent chains with 5 050 000 iterations for each analysis (i.e. model within phenotypic trait) and the first 50 000 were discarded as burn-in (Raftery and Lewis, 1992). A total of 100 000 samples of model parameters were saved from each chain with a lag interval of 50 iterations; samples from all three chains were used to characterize the posterior distribution of each model parameter invoking the ergodic property of the chains (Gilks et al, 1996).…”