|
Activity Number:
|
238
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Tuesday, July 31, 2007 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Bayesian Statistical Science
|
| Abstract - #309197 |
|
Title:
|
Density Estimation and Dimension Reduction with Logistic Gaussian Process Priors
|
|
Author(s):
|
Surya Tokdar*+
|
|
Companies:
|
Carnegie Mellon University
|
|
Address:
|
5000 Forbes Ave, Pittsburgh, PA, 15213,
|
|
Keywords:
|
Markov chain Monte carlo ; L-1 convergence ; Smoothing ; Bayesian semiparametric modeling
|
|
Abstract:
|
Logistic Gaussian process priors provide an extremely flexible framework to build smooth, nonparametric models for density estimation. Lack of efficient computing methods, however, have kept these priors somewhat underused and understudied. We propose a new method that allows efficient computation with these priors for density estimation. This method can be easily generalized to other applications with models based on transforms of a Gaussian process. We illustrate this with an application to sufficient dimension reduction in multiple regression. In either application we prove that strong posterior consistency properties obtain for a rich class of models.
|
