Adjustment of normal mixed models through triple minimization

by Dário Ferreira, Sandra Ferreira Célia Nunes and João Tiago Mexia.

Abstract: The technique of Triple minimization reduces the problem of estimation of the parameters of normal mixed models to the estimation of the relative variance components. Thus, since the square of the norm of these variance components is equal to 1, the solution will belong to the set of vectors with norm 1 and non-negative components. This fact will be used in this paper to carry out the estimation of the relative variance components through stochastic search. The remaining parameters are then estimated.

Key Words: Triple minimization, normal mixed models, inference, stochastic search

Authors:
Dário Ferreira, dario@mat.ubi.pt
Sandra Ferreira, sandra@mat.ubi.pt
Célia Nunes, celia@mat.ubi.pt
João Tiago Mexia, jtm@fct.unl.pt

Editor: Fellman, Johan O., fellman@hanken.fi

READING THE ARTICLE: You can read the article in portable document (.pdf) format (122546 bytes.)

NOTE: The content of this article is the intellectual property of the authors, who retains all rights to future publication.

This page has been accessed 3325 times since SEPTEMBER 24, 2009.


Return to the InterStat Home Page.