A Typology of Outcome Patterns in Three-Variable Models

Abstract Developing a lexicon of variable types is essential for effective research, but existing variable typologies use different terms for the same process, and remain incomplete. We offer a typology in which extraneousness is distinguished from mediation and moderation; suppression is subsumed under mediation; and the hypothesized role of X or Z in a three-variable system can be recast. Simple multiple regression and path analytic examples are used to clarify the difference between extraneousness, moderation, and mediation, and suppressive mediation. The practical and theoretical importance of mediation, including a rule-of-thumb for identifying its relevance, is addressed.