Sankhya: The Indian Journal of Statistics

1993, Volume 55, Series A, Pt. 3, 506--515

NUMERICAL PROCEDURES FOR ESTIMATING THE PARAMETERS IN A MULTIVARIATE HOMOGENEOUS CORRELATION MODEL WITH UNEQUAL VARIANCES

By

JUAN C. MEZA, Sandia National Laboratories, U. S. A.

And

INGRAM OLKIN, Stanford University, U. S. A.

 

SUMMARY. Closed form expressions for the maximum likelihood estimates (MLE) of the parameters in a multivariate normal distribution in which the variances are homogeneous and the correlations are equal is well-known. However, when the correlations are equal but the variances are not homogeneous, no closed form expressions are available. We provide two iterative procedures that converge rapidly. One procedure uses an extension of a well known scaling method, which is itself of intrinsic interest.

AMS Subject classification. 62H12

FULL PAPER.

This article in Mathematical Reviews.