We propose to combine two quite powerful ideas that have recently appeared in the Markov chain Monte Carlo literature: adaptive Metropolis samplers and delayed rejection. The ergodicity of the resulting non-Markovian sampler is proved, and the efficiency of the combination is demonstrated with various examples. We present situations where the combination outperforms the original methods: adaptation clearly enhances efficiency of the delayed rejection algorithm in cases where good proposal distributions are not available. Similarly, delayed rejection provides a systematic remedy when the adaptation process has a slow start.
KeywordsAdaptive Markov chain Monte Carlo Adaptive Metropolis-Hastings Delayed rejection Efficiency ordering
Unable to display preview. Download preview PDF.
- Andrieu C. and Robert C.P. 2001. Controlled MCMC. Preprint.Google Scholar
- Andrieu C. and Moulines E. 2002. On the ergodicity properties of some adaptive MCMC algorithms. To appear in Annals of Applied Probability.Google Scholar
- Bowie G.L., Mills W.B., et al. 1985. Rates, constants, and kinetic formulations in surface water modeling. Technical Report EPA/600/3-85/040, U.S. Environmental Agency, ORD, Athens, GA, ERL.Google Scholar
- Gelman A.G., Roberts G.O., and Gilks W.R. 1996. Efficient Metropolis jumping rules. In: Bernardo J.M., Berger J.O., David A.F., and Smith A.F.M. (Eds.), Bayesian Statistics V. Oxford University Press, pp. 599–608.Google Scholar
- Malve O., Laine M., Haario H., Kirkkala T., and Sarvala J. Bayesian modeling of algae mass occurrences—using adaptive MCMC methods with a lake water quality model. To appear in Environmental Modelling and Software, 2006.Google Scholar
- Mira A. 2001. On Metropolis-Hastings algorithms with delayed rejection. Metron, Vol. LIX, (3–4): 231–241.Google Scholar
- Sokal A.D. 1998. Monte carlo methods in statistical mechanics: Foundations and new algorithms. Cours de Troisième Cycle de la Physique en Suisse Romande. Lausanne.Google Scholar