ABSTRACT
Suppressor situations occur when the simultaneous inclusion of two predictors improves one or both validities. A common allegation is that suppressor effects rarely replicate and have little substantive import. We present substantive examples from two established research domains to counter this skepticism. In the first domain, we show how measures of guilt and shame act consistently as mutual suppressors: Adding shame into a regression equation increases the negative association between guilt and aggression, whereas adding guilt increases the positive association between shame and aggression. In the second domain, we show how the effects of self-esteem and narcissism operate consistently as mutual suppressors: That is, adding narcissism into a regression equation increases the negative association between self-esteem and antisocial behavior, whereas adding self-esteem increases the positive association between narcissism and antisocial behavior. Discussion addresses the different implications for suppressors in theoretical and variable selection applications.
