Partial Omega Squared for Anova Designs

Abstract

The present paper is concerned with measuring the size of an effect for fixed effects factorial analysis of variance (ANOVA) experimental designs. A brief review of the literature is provided, emphasizing the need to use such a measure in actual research. Measuring strength of effect is discussed in correlational terms, taking advantage of the linear model formulation for fixed effects ANOVA. It is concluded that a squared partial correlation between factor and response (partialing out effects of other factors) is usually to be preferred to the corresponding un-partialled measure (ω2) advocated by Hays (1963) and others. Examples are provided to illustrate the practical implications of this distinction.