|
Activity Number:
|
548
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Thursday, August 6, 2009 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Bayesian Statistical Science
|
| Abstract - #303801 |
|
Title:
|
Center-Adjusted Inference for a Nonparametric Bayesian Random Effect Distribution
|
|
Author(s):
|
Yisheng Li*+ and Peter Mueller and Xihong Lin
|
|
Companies:
|
The University of Texas M.D. Anderson Cancer Center and The University of Texas M.D. Anderson Cancer Center and Harvard School of Public Health
|
|
Address:
|
1400 Pressler St., Houston, TX, 77030,
|
|
Keywords:
|
Bayesian nonparametric model ; fixed effect ; generalized linear mixed model ; moments of a Dirichlet process ; post processing ; random probability measure
|
|
Abstract:
|
Dirichlet process (DP) prior models are a popular choice for semiparametric Bayesian random effect models. The base measure in the DP is often specified as a normal distribution with mean zero. The fact that this prior allows the random effect distribution to have a non-zero mean complicates interpretation of the inference for the fixed effects paired with the random effects. The resulting inference on the fixed effects can be biased and poor. We propose a post-processing technique to adjust the inference. The adjustment is based on an analytic evaluation of the random moments of the DP and has no additional computational cost. We evaluate its performance using simulations and apply it to a prostate specific antigen (PSA) data set. We provide an R function that allows investigators to implement the proposed adjustment in a post-processing step without changing the simulation itself.
|
