|
Activity Number:
|
123
|
|
Type:
|
Invited
|
|
Date/Time:
|
Monday, August 7, 2006 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Bayesian Statistical Science
|
| Abstract - #305212 |
|
Title:
|
Recent Advances in Posterior Convergence Problem
|
|
Author(s):
|
Subhashis Ghosal*+
|
|
Companies:
|
North Carolina State University
|
|
Address:
|
Department of Statistics, Raleigh, NC, 27695-8203,
|
|
Keywords:
|
posterior ; consistency ; convergence rate ; infinite dimensional model
|
|
Abstract:
|
The study of posterior consistency and convergence rates of Bayes procedures in infinite dimensional models is challenging. Counterexamples show convergence may fail fairly easily. Conditions ensuring consistency were known from the mid-60s under the i.i.d. set-up, and were generalized in the 90s. Lately, these ideas have been extended to study the posterior convergence rates in the infinite dimensional setting. Attention has been given to generalizing the results to incorporate non-i.i.d. observations. We will present results in that direction. Recently, an alternative approach to posterior convergence via a martingale property of integrated likelihood was considered in the literature. We also will discuss relations of some of our results with the ones obtained by the martingale approach.
|
