Institute of Statistics and Decision Sciences

Duke University

presents:

Alex Reutter

Institute of Statistics & Decision Sciences, Duke University

"General Strategies for Assessing Convergence of MCMC Algorithms Using Coupled Sample Paths"

Abstract:

In a previous article, Johnson (1996) proposed a coupling algorithm to study convergence properties of certain Markov chain Monte Carlo (MCMC) algorithms. We extend this method to arbitrary MCMC sampling schemes.

Our procedures rely on Markov-maximal coupling or mixture coupling algorithms to update multiple MCMC sample paths initialized from independent draws from an overdispersed estimate of the target distribution. By observing the distribution of coupling iteration, we obtain bounds on the total variation distance of iterates in a chain from their target distribution.

Wednesday, September 27, 1995

11:45 - 12:45

116 Old Chemistry Building

Any questions concerning the seminar may be addressed to Cheryl McGhee @ (919) 684-8029, e-mail cheryl@isds.duke.edu, or finger seminar@isds.duke.edu.