|
Activity Number:
|
566
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Thursday, August 6, 2009 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Nonparametric Statistics
|
| Abstract - #304569 |
|
Title:
|
Bayesian Empirical Likelihood for Quantile Regression
|
|
Author(s):
|
Yunwen Yang*+
|
|
Companies:
|
University of Illinois at Urbana-Champaign
|
|
Address:
|
725 South Wright Street, Champaign, IL, 61820,
|
|
Keywords:
|
Bayesian inference ; empirical likelihood ; quantile regression
|
|
Abstract:
|
Empirical likelihood has been shown to share many of the asymptotic properties as parametric likelihood. We consider the possibility of incorporating empirical likelihood in Bayesian inference for quantile regression models. This framework allows joint inference on multiple quantiles without a parametric likelihood. We provide an MCMC strategy along with a modified Newton's algorithm for the empirical likelihood calculation, and discuss the advantages and challenges in the Bayesian empirical likelihood approach to quantile estimation. We show through simulation that the Bayesian empirical likelihood approach adds value to quantile estimation, and verify in theory that the posterior distributions have desirable asymptotic properties. The talk is based on joint work with Professor Xuming He.
|
