Skip Navigation

Biometrika 1999 86(1):153-167; doi:10.1093/biomet/86.1.153
© 1999 by Biometrika Trust
This Article
Right arrow Full Text (PDF)
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Alert me to new issues of the journal
Right arrow Add to My Personal Archive
Right arrow Download to citation manager
Right arrowRequest Permissions
Google Scholar
Right arrow Articles by Fernandez, C
Right arrow Articles by Steel, M.
Right arrow Search for Related Content
Social Bookmarking
Add to CiteULike   Add to Connotea   Add to Del.icio.us  
What's this?

Multivariate Student-t regression models: Pitfalls and inference

C FernandezA1 and MFJ SteelA2

A1 Department of Mathematics, University of Bristol, Bristol, BS8 1TW, UK carmen.fernandez@bristol.ac.uk A2 Department of Economics, University of Edinburgh, Edinburgh EH8 9JY, UK mark.steel@ed.ac.uk

We consider likelihood-based inference from multivariate regression models with independent Student-t errors with unknown degrees of freedom. Some pitfalls are revealed of both Bayesian and maximum likelihood methods. Under a commonly used non-informative prior, Bayesian inference is precluded for certain samples, even though a well-defined conditional distribution exists of the parameters given the observables. We also find that adding new observations can destroy the possibility of conducting posterior inference. Global maximisation of the likelihood function is a vacuous exercise since the latter is unbounded as we tend to the boundary of the parameter space. The unboundedness of the likelihood function also implies that the problems mentioned above for Bayesian analysis can even occur under proper priors. These pitfalls arise as a consequence of the fact that the recorded data have zero probability under the assumed sampling model. Therefore, a Bayesian analysis on the basis of set observations, which takes into account the precision with which the data were originally recorded, i.e. the 'rounding', is proposed and illustrated by several examples.

Keywords:Bayesian inference; Continuous distribution; Maximum likelihood; Missing data; Scale mixture of normals. of


Add to CiteULike CiteULike   Add to Connotea Connotea   Add to Del.icio.us Del.icio.us    What's this?


This article has been cited by other articles:


Home page
 BiometrikaHome page
T. C. O. Fonseca, M. A. R. Ferreira, and H. S. Migon
Objective Bayesian analysis for the Student-t regression model
Biometrika, June 1, 2008; 95(2): 325 - 333.
[Abstract] [PDF]

Home page
 Statistical ModellingHome page
C. Savalli, G. A Paula, and F. J. Cysneiros
Assessment of variance components in elliptical linear mixed models
Statistical Modeling, April 1, 2006; 6(1): 59 - 76.
[Abstract] [PDF]

Home page
 Statistical ModellingHome page
J. Taylor and A. Verbyla
Joint modelling of location and scale parameters of the t distribution
Statistical Modeling, July 1, 2004; 4(2): 91 - 112.
[Abstract] [PDF]


Disclaimer:
Please note that abstracts for content published before 1996 were created through digital scanning and may therefore not exactly replicate the text of the original print issues. All efforts have been made to ensure accuracy, but the Publisher will not be held responsible for any remaining inaccuracies. If you require any further clarification, please contact our Customer Services Department.