An Efficient Parallel Simulation Method for Posterior Inference on Paths of Markov Processes

Abstract

In this note, we propose a method for efficient simulation of paths of latent Markovian state processes in a Markov Chain Monte Carlo setting. Our method harnesses available parallel computing power by breaking the sequential nature of commonly encountered state simulation routines. We offer a worked example that highlights the computational merits of our approach.

Keywords: Bayesian inference, Markov Chain Monte Carlo, Posterior path simulation

JEL Classification: C11, C15