|
Activity Number:
|
278
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Tuesday, August 4, 2009 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Bayesian Statistical Science
|
| Abstract - #304516 |
|
Title:
|
Hierachical Bayesian Nonparametric Approaches for Detecting Difference Boundaries in Disease Maps
|
|
Author(s):
|
Pei Li*+ and Sudipto Banerjee and Tim Hanson and Marshall A. McBean
|
|
Companies:
|
The University of Minnesota and The University of Minnesota and The University of Minnesota and The University of Minnesota
|
|
Address:
|
, , ,
|
|
Keywords:
|
Difference boundaries ; Dirichlet Process ; Markov random fields
|
|
Abstract:
|
In many applications involving spatially referenced data in public health, investigators want to understand the underlying mechanisms causing disparities in the outcome variables. Statistical methods correctly accounting for uncertainty at various levels to elicit spatial zones of rapid change help epidemiologists and policy-makers better understand the factors driving these disparities. We will discuss Bayesian nonparametric approaches using a really dependent stick-breaking priors to achieve fully model-based inference on regions to reveal clusters or boundaries that suggest hidden risk factors. In particular, we concentrate on models that embed univariate and multivariate Markov random fields within a nonparametric framework. We illustrate with simulated data as well as Pneumonia and Influenza hospitalization data from the SEER-Medicare program in Minnesota.
|
