Alternative Formulations and Procedures for the Two-Way Mixed Model

Publisher Website
Google Scholar
Abstract
Various alternative formulations for the 2-way mixed model can be used interchangeably in analyzing data, provided that certain relationships between the formulations with respect to the variance components and the random effects are recognized and that certain differences between the parameter spaces do not come into play. The balanced 2-way classification is used to illustrate how Henderson''s mixed-model procedures for estimating linear functions of fixed and random effects differ from procedures that treat all effects as fixed. Explicit expressions are obtained for the balanced 2-way model, for variance-component estimators yielded by each of 3 procedures: maximum likelihood (ML), restricted maximum likelihood (REML) and a pseudo-Bayesian modification of REML. Mean squared error comparisons among these 3 procedures tend to favor the pseudo-Bayesian procedure.

This publication has 2 references indexed in Scilit: