Mediation, Moderation & Interaction: Definitions, Discrimination & (Some) Means of Testing
This is the author accepted manuscript. The final version is available from Sense Publishers via the DOI in this record.
In 1986 Baron and Kenny set out to clarify the terms “Mediation” and “Moderation” as used in the social sciences (with the origins of each described by Roe, 2012). Twenty six years later, the seminal paper that this collaboration resulted in (Baron & Kenny, 1986) has been cited over 35,000 times (35,672 via Google Scholar as of 09/01/2013). However, despite this extensive record of citation, uncertainty continues to surround the use of these terms in social science research and they have received relatively little attention in specifically educational research (cf. Kraemer, Stice, Kazdin, Offord, & Kupfer, 2001). Partly in response to this uncertainty, and partly in response to advances made in the application of more complex statistical analyses in educational research (e.g. Creemers, Kyriakides, & Sammons, 2010; Goldstein, 2003; Luyten & Sammons, 2012; Tatsuoka, 1973), this chapter is made-up of four sections which together provide the quantitative educational researcher with an up to date understanding of these terms as well as examples of their current implementation to test theoretical models and address notions of causality. These four sections are titled: 1) Unambiguous Definitions 2) Discriminating Mediation, Moderation, and Interaction 3) Some means of testing Mediation and Moderation 4) Testing Moderation: An example through three equivalent statistical analyses Together, the first two sections of this chapter present simple, clear definitions that distinguish “Mediation”, “Moderation”, and “Interaction” both from each other as well as from a number of other commonly-used terms. Section 3 then presents a number of statistical methods by which these terms can be statistically operationalised. This third section pays particular attention to Moderation as the statistical methods associated with it (in comparison to Mediation) are particularly varied and numerous. The final section of this paper (Section 4) then builds upon the focus on Moderation within Section 3 by presenting an example Moderation from educational research conducted within the early years (for children under age 5 years) which is then statistically operationalised and tested by three equivalent parallel analyses.
In Teo, Timothy (ed) Handbook of Quantitative Methods for Educational Research, Sense Publishers, 2013, pp. 267 - 286