The Simulation Smoother for Time Series Models

This paper introduces a simple and efficient method, the simulation smoother, for sampling from the smoothing density associated with time series models. The simulation smoother exploits the common structure which most time series models possess. Simulation smoothing comes up in a number of papers, e.g. FrühwirthSchnatter (1994, 1995), Carter and Kohn (1994), Shephard (1994), Chib and Greenberg (1994), and may be of importance in future time series work. One area of application is the Markov chain Monte Carlo method or Gibbs sampling, described by Smith and Roberts (1993) and Ripley (1987, pp. 113–6).