Contents

CHAPTER 6: Independent Groups Designs 205Key Conceptin an experiment. The mean difference between two groups is always relativeto the average variability in participants’ scores. One frequently used measureof effect size is Cohen’s d. Cohen (1992) developed procedures that are nowwidely accepted. He suggested that d values of .20, .50, and .80 represent small,medium, and large effects of the independent variable, respectively.We can illustrate the use of Cohen’s d as a measure of effect size by comparingtwo conditions in the video-game experiment, the reward condition andthe nonviolent condition. The d value is .83 based on the difference between themean aggressive cognition in the reward condition (.210) and the nonviolentcondition (.157). This d value allows us to say that the video-game independentvariable, reward vs. nonviolent, had a large effect on the aggressive cognitionin these two conditions. Effect-size measures provide researchers with valuableinformation for describing the findings of an experiment.Measures of central tendency and variability, as well as effect size, aredescribed in Chapters 11 and 12. In those chapters we outline the computationalsteps for these measures and discuss their interpretation. Manydifferent effect-size measures are found in the psychology literature. In additionto Cohen’s d, for example, a popular measure of effect magnitude iseta squared, which is a measure of the strength of association between theindependent and dependent variables (see Chapter 12). That is, eta squaredestimates the proportion of total variance accounted for by the effect of theindependent variable on the dependent variable. Measures of effect size aremost helpful when comparing the numeric values of a measure from two ormore studies or when averaging measures across studies, as is done when ameta-analysis is performed (see below).Key ConceptResearchers also use measures of effect size in a procedure called metaanalysis.Meta-analysis is a statistical technique used to summarize the effectsizes from several independent experiments investigating the same independentor dependent variable. In general, the methodological quality of the experimentsincluded in the meta-analysis will determine its ultimate value (seeJudd, Smith, & Kidder, 1991). Meta-analyses are used to answer questionslike: Are there gender differences in conformity? What are the effects of classsize on aca demic achievement? Is cognitive therapy effective in the treatmentof depression? Box 6.2 describes a meta-analysis of studies on effective psychotherapyfor youth with psychological disorders. The results of individualexperiments, no matter how well done, often are not sufficient to provide answersto questions about such important general issues. We need to consider abody of literature (i.e., many experiments) pertaining to each issue. (See Hunt,1997, for a good and readable introduction to meta-analysis.) Meta-analysisallows us to draw stronger conclusions about the principles of psychologybecause these conclusions emerge only after looking at the results of manyindividual experiments. These analyses provide an efficient and effective wayto summarize the results of large numbers of experiments using effect-sizemeasures.