BAYES' THEOREM AND THE USE OF PRIOR KNOWLEDGE IN REGRESSION ANALYSIS
Abstract
A Bayesian approach to the problem of integration prior information into the analysis of the normal regression model was adopted. A reinterpretation of the ''mixed'' estimaticedure of Theil and Goldberger was provided with an assumption. It was shown that the posterior distribution of beta takes the form of a product of multivariate normal and multivariate t distributions. What may be regarded as a generalization of Fisher's work on the problem of making inferences when samples are drawn from two normal populations with common mean and unequal variances was obtained. In this case, it was shown that the posterior dis tribution of beta is in the form of the product of two multivariate t distributions.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jan 01, 1963
- Accession Number
- AD0409435
Entities
People
- Arnold Zellner
- George C. Tiao
Organizations
- University of Wisconsin–Madison