Contents

260 PART III: Experimental Methodsmethod can be used only when one of the independent variables has two levels.For complex designs when both independent variables have three or morelevels, graphs should be used to identify interaction effects.Complex Designs with Three Independent VariablesThe power and complexity of complex designs increase substantially when thenumber of independent variables in the experiment increases from two to three.In the two-factor design there can be only one interaction effect, but in the threefactordesign each independent variable can interact with each of the other twoindependent variables and all three independent variables can interact together.Thus, the change from a two-factor to a three-factor design introduces the possibilityof obtaining four different interaction effects. If the three independentvariables are symbolized as A, B, and C, the three-factor design allows a testof the main effects of A, B, and C; two-way interaction effects of A B, A C,B C; and the three-way interaction effect of A B C. The efficiency of anexperiment involving three independent variables is remarkable. An experimentinvestigating discrimination in the workplace will give you a sense of just howpowerful complex designs can be.Pingitore, Dugoni, Tindale, and Spring (1994) investigated possible discriminationagainst moderately obese people in a mock job interview. Participantsin the experiment viewed videotapes of job interviews. In one of their experimentsthey used a 2 2 2 design. The first independent variable was theweight of the applicant (normal or overweight). The role of the applicant forthe job in the videotapes was played by professional actors who were of normalweight. In the moderately obese conditions, the actors wore makeup andprostheses so that they appeared 20% heavier. The second independent variablein the experiment was the sex of the applicant (male or female). The thirdindependent variable was participants’ concern about their own body and theimportance of body awareness to their self-concept (high or low). This variablewas defined using a self-report measure of how participants viewed their body.A natural groups design was used to study this “body-schema variable.” Participantswere randomly assigned to evaluate male or female applicants who werenormal weight or moderately obese (random groups designs). The dependentvariable was the participants’ rating on a 7-point scale of whether they wouldhire the applicant (1 definitely not hire and 7 definitely hire).The results of the Pingitore et al. experiment for these three variables areshown in Figure 8.5. As you can see, displaying the means for a three-variableexperiment requires a graph with more than one “panel.” One panel of the figureshows the results for two variables at one level of the third variable, and theother panel shows results for the same two variables at the second level of thethird independent variable.As you are now familiar with main effects and simple (two-way) interactioneffects, let us concentrate on understanding a three-factor or three-way interactioneffect. As you can see in Figure 8.5, a two-way interaction effect of theapplicant’s weight and sex occurred only with participants who were high inconcern about their own bodies. That is, those high on the body-schema variable