Antithetic coupling of two Gibbs sampler chains

Arnoldo Frigessi*, Jørund Gåsemyr, Håvard Rue

*Corresponding author for this work

Research output: Contribution to journalArticlepeer-review

20 Scopus citations

Abstract

Two coupled Gibbs sampler chains, both with invariant probability density π, are run in parallel so that the chains are negatively correlated. We define an asymptotically unbiased estimator of the π-expectation E(f(X)) which achieves significant variance reduction with respect to the usual Gibbs sampler at comparable computational cost. The variance of the estimator based on the new algorithm is always smaller than the variance of a single Gibbs sampler chain, if π is attractive and f is monotone nondecreasing in all components of X. For nonattractive targets π, our results are not complete: The new antithetic algorithm outperforms the standard Gibbs sampler when π is a multivariate normal density or the Ising model. More generally, nonrigorous arguments and numerical experiments support the usefulness of the antithetically coupled Gibbs samplers also for other nonattractive models. In our experiments the variance is reduced to at least a third and the efficiency also improves significantly.

Original languageEnglish (US)
Pages (from-to)1128-1149
Number of pages22
JournalAnnals of Statistics
Volume28
Issue number4
DOIs
StatePublished - Aug 2000
Externally publishedYes

Keywords

  • Antithetic Monte Carlo
  • Associated random variables
  • Attractive models
  • Decay of cross-autocorrelations
  • Markov chain Monte Carlo
  • Variance reduction

ASJC Scopus subject areas

  • Statistics and Probability
  • Statistics, Probability and Uncertainty

Fingerprint

Dive into the research topics of 'Antithetic coupling of two Gibbs sampler chains'. Together they form a unique fingerprint.

Cite this