November 18: Richard Tweedie, University of Minnesota

UNIVERSITY OF MINNESOTA

SEMINAR

School of Statistics
College of Liberal Arts

Perfect Sampling and Harris Chains

Richard Tweedie
Division of Biostatistics
University of Minnesota

Thursday, November 18, 1999
4:00-5:00 PM, Room 2-215 Carlson School of Management
Social at 3:30 PM in Room 531 Heller Hall (formerly Management/Economics)

Abstract

In this talk we introduce several implementations of the Propp-Wilson algorithm for for simulating ``perfect'' random samples from the invariant measure of a recurrent Markov chain. For a Harris chain on a continuous space, we describe a new method which uses backward coupling of embedded regeneration times, and wor ks most effectively for finite chains, or on continuous spaces for stochastical ly monotone chains, where paths may be sandwiched between ``upper'' and ``lower'' processes.

Examples show that more naive approaches to constructing such bounding processes may be considerably biased, but that the algorithm can be simplified in certain cases to make it easier to run. We give explicit analytic bounds on the backwar d coupling times in the stochastically monotone case. An application of the algorithm to storage models is given.