|
Activity Number:
|
299
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Wednesday, August 11, 2004 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Bayesian Statistical Science
|
| Abstract - #301876 |
|
Title:
|
Fitting Response Time Models by Adaptive Importance Sampling
|
|
Author(s):
|
Cheryl Niermann*+ and Mario Peruggia and Trisha Van Zandt
|
|
Companies:
|
Ohio State University and Ohio State University and Ohio State University
|
|
Address:
|
227 Cockins Hall, Columbus, OH, 43210,
|
|
Keywords:
|
Bayesian modeling ; MCMC algorithms ; mixture distributions ; shifted exponential distribution ; shifted Weibull distribution
|
|
Abstract:
|
Due to the growing popularity of Bayesian models in the social sciences, there is a need to develop user-friendly computational tools to perform model-fitting and model assessment tasks. A software environment like BUGS allows a user to perform MCMC estimation on the basis of approximate posterior draws. The user need only specify the Bayesian model in a simple and intuitive manner without having to write specialized computer code to implement the MCMC algorithm. However, the choice of models that can be fit is limited by the list of distributions allowed by BUGS. In our analyses of response time data arising from cognitive experiments we used importance sampling to reweight MCMC output from BUGS and fit models whose likelihoods cannot be handled directly. This talk will describe the models that we fit, will outline the strategies that we used to construct and adapt sequentially the importance sampling densities so as to improve the accuracy of our inferences, and will illustrate the performance of the methodology on simulated and real data.
|
