Abstract
Moderated regression models include an interaction, or product term, and can be used to assess whether the relationship between a given independent variable (IV) and a dependent variable (DV) depends on a third moderator variable (MV). If the moderation effect is significant, researchers recommend either ignoring main effects completely, or carefully interpreting them as conditional effects. However, when the moderation effect is not significant, this implies that the typical interpretation of main effects as average effects is appropriate. The present study challenges this claim since lack of significance may be due to lack of power rather than to no true population effect. To explore this idea, a simulation study is conducted and analytic illustration provided. Results indicate that when a true moderation effect exists, it may not be detected, implying the potential for misleading interpretation of main effects. To guard against this, applied researchers are encouraged to conduct power analyses prior to a moderation study; to mean-center predictors; to consider exploring the main-effects-only model by omitting the interaction effect; and to consider information criteria approaches to testing effects.