American Statistical Association
New York City
Metropolitan Area Chapter

New York State Psychiatric Institute
at Columbia University Medical Center
Biostatistics Seminar



COMPUTATION IN BAYESIAN MODELS FOR SURVEY SCALE USAGE

by

Christopher Hans, Ph.D.
Department of Statistics
The Ohio State University


Abstract

A common modeling approach for data collected on discrete scales is to introduce continuous latent variables, treated as missing data that are linked to the observations via a censoring mechanism. Bayesian approaches typically use data augmentation within a Markov chain Monte Carlo (MCMC) algorithm that simplifies analysis by avoiding direct evaluation of the likelihood. However, if not carefully implemented, the cost of data augmentation is that Markov chain mixing can be severely degraded, rendering the method useless for inference. Motivated by complex, high-dimensional models for survey scale-usage data, we introduce a covariance decomposition for the analysis that greatly facilitates Bayesian model fitting. We demonstrate that our model fitting method is fast and accurate using a customer satisfaction survey dataset. We also discuss extensions of the model that are appropriate for another dataset based on a survey of hospital patients.

Biographical Note

Christopher Hans is an Assistant Professor in Department of Statistics at The Ohio State University. He received his Ph.D. in Statistics from Duke University. Dr. Hansí main research interest is in the development of Bayesian methodology for the analysis of modern, complex datasets. He is particularly interested in the development of related computational methods. Other areas of interest include Markov chain Monte Carlo methods as well as the use of parallel computing in statistics. Specific methodological research areas include the problems of variable selection and model uncertainty in contexts of regression, prediction, and complex multivariate modeling with many variables. A key element of his research is the development of stochastic search and MCMC methods for exploring large model spaces. Recent focus has been on the development of new classes of prior distributions for regression problems that provide connections to penalized optimization procedures.


Date: Tuesday, May 4, 2010
Time: 3:00 - 4:00 P.M.
Location: New York State Psychiatric Institute
1051 Riverside Drive
6th Floor Multipurpose Room (6602)
New York, New York
(Directions)

RESERVATIONS ARE NOT REQUIRED

Coffee: 2:45 to 3:00 P.M.
Reception: 4:00 to 4:30 P.M.


Home Page | Chapter News | Chapter Officers | Chapter Events
Other Metro Area Events | ASA National Home Page | Links To Other Websites
NYC ASA Chapter Constitution | NYC ASA Chapter By-Laws

Page last modified on April 29, 2010

Copyright © 1998-2010 by New York City Metropolitan Area Chapter of the ASA
Designed and maintained by Cynthia Scherer
Send questions or comments to nycasa@mindspring.com