Wavelet-Based Power Transformation for Long Memory Regression Models with non-Gaussian Errors

• Activity Number: 302
• Type: Contributed
• Date/Time: Tuesday, August 3, 2010 : 8:30 AM to 10:20 AM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Kyungduk Ko*+
• Companies: Boise State University
• Keywords: Long memory ; Box-Cox transform ; Gibbs sampler

Abstract

We consider a linear regression model with its response variable being non-Gaussian and long range dependent, and perform the Box-Cox transform for achieving approximate normality of the response variable. To this end we propose a wavelet-based estimation method that provides a simultaneous estimation of the model parameters under the assumption of unknown transformation and long memory parameters, a situation which arises commonly in practice. Fractional Gaussian noise and I(d) process are considered for the long range dependent error. We explore a posterior estimation method via a collapsed Gibbs sampler in the wavelet domain. Performances are assessed using simulation studies.