Appropriate moderated regression and inappropriate research strategy: A demonstration of information loss due to scale coarseness

Paunonen and Jackson (1988) demonstratedthat stepwise moderated regression provides a testof interaction effects that protects the nominalType I error rate. However, the stepwise procedurehas also been characterized as failing to detectinteraction effects in empirical studies. This issuehas led to questions regarding the method’sstatistical power (Bobko, 1986; Zedeck, 1971) inapplied research. It is demonstrated that, becauseof a research strategy frequently used in empiricalinvestigations, the probability of Type II error indetecting a true interaction effect is unknown.Specifically, the number of scale steps used inmeasuring the dependent variable is shown toresult in a form of systematic error that canspuriously increase or decrease the expected effectsize of the interaction. The problem is also discussedin the context of testing more complex models.Recommendations for eliminating this problem infuture research designs are provided. Index terms:information loss, interaction effects, Likert scales,moderated regression, response transformation.