This research was aimed at evaluating and identifing which type of sum of squares can be more appropriate to test hypotheses and also presenting appropriate alternatives to solution of problems through the analysis of mean square expected values used in the methodology of mixed linear models. The analysis of mean square expected values can be a tool of great importance in analysis of data as incomplete (empty casela) as unbalanced experiment. Therefore, four examples were used each one with its pecualiarity concerning the complete or incomplete experiment with balanced or unbalanced data and in the presence of empty casela. The SAS statistical package, version Learning Edition, was used to analyze the experiments. The result of the analysis of mean square expected values indicated that the sum of squares of the type ‘I’ can be used only at of condition of completely balanced data. These results indicated on the other hand, that the sum of squares of the type ‘III’ is the most appropriate type for unbalanced data. The sum of squares of the type ‘II’ and ‘IV’ are the most important in the case of empty caselas; fact that supports the idea of a necessity of always evaluating the mean square expected values.
Avena sativa; analysis of variance; mixed models
