ABSTRACT
In this chapter we consider the statistical analysis of two groups of participants on several dependent variables simultaneously; focusing on cases where the variables are correlated and share a common conceptual meaning. That is, the dependent variables considered together make sense as a group. For example, they may be different dimensions of self-concept (physical, social, emotional, academic), teacher effectiveness, speaker credibility, or reading (blending, syllabication, comprehension, etc.). We consider the multivariate tests along with their univariate counterparts and show that the multivariate two-group test (Hotelling’s T 2) is a natural generalization of the univariate t test. We initially present the traditional analysis of variance approach for the two-group multivariate problem, and then later briefly present and compare a regression analysis of the same data. In the next chapter, studies with more than two groups are considered, where multivariate tests are employed that are generalizations of Fisher’s F found in a univariate one-way ANOVA. The last part of this chapter (sections 4.9–4.12) presents a fairly extensive discussion of power, including introduction of a multivariate effect size measure and the use of SPSS MANOVA for estimating power.
