Regenerative Simulation of Harris Recurrent Markov Chains

P. W. Glynn

Technical Report, Department of Operations Research, Stanford University (1982)

If the steady-state simulation problem associated with a general discrete-event simulation is well-posed, then the corresponding Markov chain is Harris recurrent. For Harris chains, it is possible to develop a simulation methodology, closely related to the regenerative method, for obtaining confidence intervals associated with estimation of steady-state parameters. The passage-time problem for Harris chains is also studied, and an estimation approach for the steady-state simulation problem is outlined. Finally, it is shown that a certain family of nonlinear storage processes is Harris recurrent--the family provides an example of a Harris chain for which the classical regenerative method is inapplicable.