Nonnegative Estimates of Variance Components in. Two Way Classification Models with Random Effects

Two way classification models with random effects are used widely. The variance components in these models are often estimated by ANOVA estimates. In this article, we improve the ANOVA estimates in the sense of mean square errors by an easy way. Also we construct nonnegative estimates of variance components which have smaller mean square errors than ANOVA estimates.