All-Possible-Subsets for MANOVA and Factorial MANOVAs: Less than a Weekend Project

Manova

Multivariate analyses are conducted when a researcher has a desire to consider group differences among several dependent variables simultaneously. A MANOVA is an extension of analysis of variance (ANOVA) in that, instead of examining if a variable depends on group membership, several theorized variables are examined simultaneously to determine if those variables depend on group membership (i.e., independent variable). Thus, a MANOVA is a multivariate analysis that allows researchers to address whether scores on a set of dependent variables differ as a function of a grouping variable (e.g., intervention, gender, race). When researchers decide to conduct a MANOVA, dependent variables are theoretically or empirically related and ideally both (Weinfurt, 1995). Advantages to conducting a MANOVA versus multiple ANOVAs is that a MANOVA (a) helps control for Type I error and (b) takes into account the pattern covariation among the dependent variables. As noted by Thompson (1991), one of the most important reasons for conducting multivariate methods, even more so than limiting the inflation of Type I error rates, is that “multivariate methods best honor the reality to which the researchers is purportedly trying to generalize” (p. 80).