On Two Tests for Main Effects in A Balanced Two-Way Interactive Model

Abstract

Two methods of removing interaction in a two-way balanced design were considered. Removing interaction from the model/ data will result to a reduced model and data which consequently violates the assumptions of analysis of variance (ANOVA). To resolve this problem, a linear combination method approach was used which does not violates the assumptions of ANOVA and completely makes the presence of interaction to be zero.

Keywords: Additive effects, Homogeneity of variance, Normal distribution