Regression Models under Competing Covariance Structures: A Bayesian Perspective

Jacek Osiewalski, Mark F. J. SteelAnnals 1993

Abstract

This paper develops Bayesian approaches to deal with linear elliptical regression models that differ in the covariance structure. A pretest method based on posterior model probabilities is compared with a pooling approach, and the data density is defined as a mixture of elliptical densities with weights that are unknown discrete parameters. An example with AR (1), MA (1) or uncorrelated errors is presented as an illustration of the ideas.
BibTeX
@article{osiewalski1993regression,
  author = {Jacek Osiewalski and Mark F. J. Steel},
  title = {{Regression Models under Competing Covariance Structures: A Bayesian Perspective}},
  journal = {Annals},
  year = {1993},
}

Loading related content...

Figure