[Estimating variance components in a fixed model of variance analysis].

The partition of a variance into components in experiments with fixed effects, i.e. experiments with varieties, treatment etc., is meaningful only if "components" are defined as additive parts of a whole. This also demands that the sum of the components of variance be equal to the total variance. Such additivity of the components of variance is achieved by appropriate definitions of the components of variance. Unbiassed estimates of the additively defined components of variance are obtained by a simple correction which eliminates the influence of degrees of freedom on the calculation of mean squares. This correction is not needed for the estimation of variances of random effects.The expected values of mean squares for the partition of variance and formulas for unbiassed estimates of additively defined components of variance are demonstrated in an experiment with two fixed effects and one random effect.