Zero-Variance Markov chain Monte Carlo

The main theme of this proposal is to provide a firm methodological foundation foran efficient and general purpose variancereduction technique for Markov chain Monte Carlo (MCMC) simulation. Variancereduction will be achieved by the construction of control variates. The proposed method will be tested on a large class of Bayesian statistical model to check its performance and compare it to existing alternative variance reduction techniques. The contributions of the project are both on the methodological and computational aspects of the method proposed.