Abstract
A technique given by Burdick and Sielken (1978) for determining an exact confidence interval for a positive linear combination of the two variance components in a random one-way model is examined from a different viewpoint. It is found that their technique can sometimes be improved on with respect to degrees of freedom and with respect to the particular linear combinations for which an exact confidence interval can be obtained. In addition, our development is applicable to any mixed model having two variance components and to some special cases involving three or more variance components.

This publication has 0 references indexed in Scilit: