Estimating the Parameters of a Multivariate Exponential Distribution: Part II.
Abstract
The problem of parameter estimation for the ((2 sup k)-1)-parameter, k-dimensional multivariate exponential distribution (MVE) of Marshall and Olkin is investigated. In a previous part of this report (Part I), a (k+1)-parameter version of the MVE was studied. The simpler results and concepts expounded in Part I are here extended and generalized to the ((2 sup k)-1)-parameter case. Although the MVE contains singularities, a density with respect to a (non-Lebesgue) dominating measure is specified. The likelihood equations which result, while not explicitly solvable, possess an unique root which is the maximum likelihood estimator (MLE). An alternate estimator (INT) is derived from an intuitive principle also satisfied by the MLE. INT also arises as the first iterate in a simple iterative procedure which converges to the MLE. Estimation is illustrated for the trivariate case.
Document Details
- Document Type
- Technical Report
- Publication Date
- Oct 01, 1974
- Accession Number
- ADA004319
Entities
People
- Frank Proschan
- Pasquale Sullo
Organizations
- Florida State University