The traditional analysis of variance (ANOVA) is based on the assumptions of normality, independence of the statistical errors, and equality of the variances of the errors. Studies of the robustness of the F-test have shown that the violation of normality has little effect on inferences about the means. However, the violation of independence or equality of variances can have a serious effect on
... [Show full abstract] inferences about the means, especially if the cell sample sizes are unequal (see, for example, Scheffe (1959) or Bishop (1976). In practice, the assumption of equality of error variances seems to be often unjustified; in fact, even when the error variances are equal, the power of the F-test depends upon the unknown common variance, which renders it difficult to plan an experiment rationally. Recently, Bishop and Dudewicz (1978), (1981) developed new ANOVA procedures in the contexts of the one-way layout and higher- way layouts. Their procedures allow unequal and unknown population variances and give tests with level and power completely independent of the unknown variances. Briefly outlined is the two-way layout heteroscedastic-ANOVA (HANOVA) methodology. It is followed by case study of Lin's (1978) multiple objective firm simulation for the two-way layout ANOVA procedure. Finally, the conclusion and other possible business applications are outlined.