The purpose of this article has been to summarize various non-negative estimators of variance components - both classical as well as Bayesian, proposed in the literature during recent years, for three widely applicable random effects models. The computation of these estimators has been illustrated by examples adopted from the actual case history. The m.s.e. properties of these estimators have been investigated in the papers of Klotz et al. [14], Portnoy [16], and Sahai [21,22]. Apart from their very desirable property that they always yield non-negative estimates, many of these estimators have superior m.s.e. properties over the traditional analysis of variance estimators. Thus they provide a viable alternative to them. Although many of these estimators require complex computational procedures, with wide availability of high speed calculators and computers and declining cost of computing, the computational simplicity should no longer be a consideration and it is hoped that many of these estimators would be employed in place of the traditional ones.