Activity Number:
|
339
|
Type:
|
Invited
|
Date/Time:
|
Wednesday, August 14, 2002 : 2:00 PM to 3:50 PM
|
Sponsor:
|
Section on Statistical Computing*
|
Abstract - #300208 |
Title:
|
Mixture Transition Monte Carlo: A Generalization of MCMC
|
Author(s):
|
Monnie McGee*+
|
Affiliation(s):
|
Hunter College/City University of New York
|
Address:
|
695 Park Avenue, New York, New York, 10021, USA
|
Keywords:
|
MCMC ; simulation ; Mixture Transition Distributions
|
Abstract:
|
We propose to use a mixture transition distribution (MTD) model in place of the usual Markov Chain piece of a Monte Carlo Markov Chain to correct the mixing problem often occurring when using MCMC to explore higher dimensional, multimodal distributions. We begin by defining the MTD model and give a proof that using it in MCMC leads to the appropriate equilibrium distribution. We then present simulation results involving multi-modal and/or high dimensional distributions. We investigate how well these distributions are explored with both MCMC and MTMC. Finally, we apply the new method to a real data set, discuss our results, and mention further investigations.
|