Speaker: Dr. Rick H. Hoyle, Duke University
Department of Psychology and Neuroscience

Title:  The State of Mediation Analysis in Psychological Science: Issues and Solutions

Abstract: Since the publication of Baron and Kenny’s (1986) seminal paper on mediation and moderation in psychological research, mediation analyses have become increasingly prevalent. For example, during a recent five-year period, nearly two thirds of articles published in one of the leading social psychology journals reported results from at least one mediation analysis. Under ideal conditions, the results of mediation analyses are informative. However, under conditions typical of psychological research, they can be misleading. In this talk I review the history of mediation analysis in psychology, from early work that inferred but did not measure mediating mechanisms to recent work in which mechanisms are measured and subjected to statistical testing. I argue that, although important advances have been made in statistical tests of mediated effects when putative mediators are measured, limitations related to measurement and research design limit the inferences that can be drawn from such tests. After describing and illustrating those limitations, I propose solutions that, if followed, would dramatically reduce the number of mediation analyses reported by psychological scientists but significantly increase the accuracy and informativeness of those analyses.

Note: Graduate students are encouraged to meet and chat with Rick in Room 207 in BSB from 2-3pm.

Please RSVP to Dr. Jolynn Pek.