Online Program Home
My Program

Abstract Details

Activity Number: 427
Type: Contributed
Date/Time: Tuesday, August 2, 2016 : 2:00 PM to 3:50 PM
Sponsor: Section on Statistical Computing
Abstract #319720 View Presentation
Title: The Self-Multiset Sampler
Author(s): Weihong Huang* and Yuguo Chen and Juan Shen
Companies: and University of Illinois at Urbana-Champaign and Fudan University
Keywords: Metropolis-Hastings algorithm ; Multiset ; Data augmentation ; Multimodal distribution

The Metropolis-Hastings (M-H) algorithm is one of the most well-known Markov chain Monte Carlo (MCMC) method. However, the M-H algorithm can easily get trapped in a local mode because of the stickiness of the samples. Leman et al. (2009) proposed the multiset sampler (MSS), which helps to avoid this problem. However, there are two restrictions about the MSS. First, it requires the target distribution has two parts of parameters: the interested parameter and the nuisance parameter. Second, the MSS can only be used in the cases that the nuisance parameter is discrete with finite support or continuous on a bounded set. To solve these two restrictions, we propose a new self-multiset sampler (SMSS), which extends the MSS to distributions without nuisance parameter, along with three variants of the SMSS algorithm. We also generalize our method to distributions with unbounded or infinite support. The value of our methods is demonstrated through several examples. Numerical results show that the SMSS and its generalization have a substantial advantage in sampling multimodal distributions and a faster mixing rate compared to the ordinary M-H algorithm.

Authors who are presenting talks have a * after their name.

Back to the full JSM 2016 program

Copyright © American Statistical Association