zbMATH — the first resource for mathematics

Strategies for improving MCMC. (English) Zbl 0844.62100
Gilks, W. R. (ed.) et al., Markov chain Monte Carlo in practice. London: Chapman & Hall. 89-114 (1996).
In many applications raw Markov chain Monte Carlo (MCMC) methods, in particular the Gibbs sampler, work surprisingly well. However, as models become more complex, it becomes increasingly likely that untuned methods will not mix rapidly. That is, the Markov chain will not move rapidly throughout the support of the target distribution. Consequently, unless the chain is run for very many iterations, Monte-Carlo standard errors in output sample averages will be large. We review strategies for improving run times of MCMC and our aim is to give sufficient detail for these strategies to be implemented.
For the entire collection see [Zbl 0832.00018].

62P99 Applications of statistics
65C99 Probabilistic methods, stochastic differential equations