19 - Analysis of variance in the study of interpersonal expectations: Theory testing, interaction effects, and effect sizes

Summary

The analysis of variance is only the first step in studying the results.

(Snedecor & Cochran, 1980, p. 224)

The analysis of variance (ANOVA) is the most common analysis strategy in social psychology. In a survey of three major social psychology journals in 1978, 84% of the articles used ANOVA at least some of their results (Kenny, 1985). Given this pervasiveness, one might expect ANOVA to be a robust, efficient, and powerful analytic tool. In terms of fulfilling its scientific potential, however, ANOVA has shown itself to be a disappointing underachiever. The amount of scientific knowledge that has been gained from it falls far short of the knowledge that could have (and should have) been attained.

In this chapter, several issues regarding the optimal application of ANOVA will be discussed as they pertain to social psychological research in general and interpersonal expectation research in particular. A hypothetical study involving expectation effects and interaction outcome on interpersonal attraction will be presented and analyzed. The experimental design chosen is typical. Although the data were contrived to demonstrate important points clearly, care was taken to make them representative of what might actually occur in a genuine study. The objective of this chapter is to demonstrate that the full potential of ANOVA is realized only when it performed in an active, mindful, and precise manner.

A “failed” study

An unanticipated and surprising finding to come out of the Pygmalion study (Rosenthal & Jacobson, 1966) was that the more the control group children gained in IQ, the more they were regarded as less well adjusted, less interesting, and less affectionate by their teachers, who did not expect such improvement (Rosenthal, 1974, 1985).