|
Activity Number:
|
448
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Wednesday, August 6, 2008 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Bayesian Statistical Science
|
| Abstract - #300482 |
|
Title:
|
Wavelet-Based Bayesian Estimation of Partially Linear Regression Models with Long Memory Errors
|
|
Author(s):
|
Kyungduk Ko*+ and Leming Qu and Marina Vannucci
|
|
Companies:
|
Boise State University and Boise State University and Rice University
|
|
Address:
|
1910 University Dr., Boise, ID, 83725,
|
|
Keywords:
|
Bayesian Inference ; Long Memory ; MCMC ; Partially Linear Regression Model ; Wavelet Transforms
|
|
Abstract:
|
This paper proposes a wavelet-based Bayesian estimation method of the model parameters and the nonparametric part of partially linear regression models with long memory errors. We employ discrete wavelet transforms in order to simplify the dense variance-covariance matrix of long memory errors. For a fully Bayesian inference, we adopt a Metropolis algorithm within a Gibbs sampler for the simultaneous estimation of the model parameters and nonparametric function. We evaluate performances on simulated data and then show how the proposed model can be applied to real data, by studying Northern hemisphere temperature data which is a benchmark in long memory literature.
|
