|
Activity Number:
|
543
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Thursday, August 2, 2007 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Nonparametric Statistics
|
| Abstract - #309704 |
|
Title:
|
Bayesian Isotonic Density Regression
|
|
Author(s):
|
Lianming Wang*+ and David B. Dunson
|
|
Companies:
|
National Institute of Environmental Health Sciences and National Institute of Environmental Health Sciences
|
|
Address:
|
111 T. W. Alexander Drive, Research Triangle Park, NC, 27709,
|
|
Keywords:
|
Conditional density estimation ; Dependent Dirichlet process ; Hypothesis test ; Nonparametric Bayes ; Quantile regression ; Stochastic ordering
|
|
Abstract:
|
In estimating and performing inferences on conditional response distributions given predictors, stochastic ordering constraints can be incorporated to express prior knowledge and improve efficiency. This article proposes a nonparametric Bayes approach for modeling of an uncountable collection of stochastically ordered distributions indexed by a continuous predictor. Theory is developed to allow priors to be chosen with large support through a dependent Dirichlet process (DDP) specification. Choosing monotone splines for functional atoms within the DDP representation, an efficient MCMC algorithm is developed for posterior computation. Methods are developed for hypothesis testing and graphical presentation of results. The approach is evaluated through a simulation study and is applied to an epidemiologic study.
|
