|
Activity Number:
|
15
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Sunday, August 8, 2004 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Bayesian Statistical Science
|
| Abstract - #301052 |
|
Title:
|
Importance Reweighting within Gibbs for Hierarchical Regression Modeling
|
|
Author(s):
|
Li-Jung Liang*+ and Robert E. Weiss
|
|
Companies:
|
University of California, Los Angeles and University of California, Los Angeles
|
|
Address:
|
Biostatistics, 3161 Sepulveda Blvd., Los Angeles, CA, 90034,
|
|
Keywords:
|
hierarchical regression model ; mixture of Dirichlet processes ; Markov chain Monte Carlo ; phylogenetic ; HIV
|
|
Abstract:
|
We develop a statistical model and computational algorithm for combining analyses of a number of datasets. Individual analyses are fit independently using previously written stand-alone software that fits a complex Bayesian model using MCMC simulation. Each individual analysis is computationally intensive and MCMC output from each of these complex Bayesian analyses is available. Constructing a large complex model involving all the original datasets is time consuming and may be difficult. Instead, our strategy is to use the existing MCMC samples of the individual posteriors. We place a hierarchical regression model across the individual analyses for estimating parameters of interest within and across analyses. Our model has two key features. We use a MDP prior for the parameters of interest to relax parametric assumptions and to ensure the prior distribution for the parameters of interest is continuous. We use an importance reweighting algorithm within Gibbs to sample values of the individual parameters. We demonstrate our approach on a set of phylogenetic models of HIV-1 nucleotide sequence data.
|
