Prof. Dr. D. Rudolf
Abstract:
Perturbation theory for Markov chains addresses the question of how small differences in the transition probabilities of Markov chains are reflected in differences between their distributions. Under a convergence condition we present an estimate of the Wasserstein distance of thenth step distributions
between an ideal, unperturbed and an approximating, perturbed Markov chain. We illustrate the result with an example of an autoregressive process as well
as an application to the Monte Carlo within Metropolis algorithm
