Variance components in split-plot experiments with factorial scheme in the whole plots

Abstract

The components of variance for balanced split-plot experiments were determined, considering a complete factorial design with two factors distributed in the whole plots, and a third factor in the subplots. The model was considered to have fixed, random or mixed effects, and having for each case the best linear combination of the mean squares for the F test procedure. For the fixed model it was observed that the test should be made without modification of the usual rules for the split-plot experiments. For random or mixed models combinations of mean squares should be used to carry out F tests.