25 Statistical Mediation Analysis

Research Methods in Psychology


  1. David P. MacKinnon1,
  2. Davood Tofighi PhD2

Published Online: 26 SEP 2012

DOI: 10.1002/9781118133880.hop202025

Handbook of Psychology, Second Edition

Handbook of Psychology, Second Edition

How to Cite

MacKinnon, D. P. and Tofighi, D. 2012. Statistical Mediation Analysis. Handbook of Psychology, Second Edition. 2:IV:25.

Author Information

  1. 1

    Arizona State University, Department of Psychology, Tempe, Arizona, USA

  2. 2

    Georgia Institute of Technology, School of Psychology, Atlanta, Georgia, USA

Publication History

  1. Published Online: 26 SEP 2012


Hypotheses regarding how an independent variable affects a dependent variable via a mediating variable are widespread in both basic and applied psychology. This chapter focuses on statistical and design methods to investigate mediation relations rather than the substantive importance of mediation that is described elsewhere (MacKinnon, 2008). The chapter starts with detailed information on the single mediator model including covariance between estimates, measures of effect size, hypothesis testing, confidence limit estimation, and Bayesian methods. Causal inference approaches for mediation are described. Comprehensive mediation models are then discussed including models that accommodate both moderator and mediator variables, multiple mediators, multilevel models, and models that incorporate longitudinal relations among variables. We acknowledge that the identification of mediating variables can be a challenging process requiring a variety of information in addition to statistical analysis such as replication and experimental studies. Although the identification of mediating variables is a challenging task, many new statistical and methodological tools have been developed to help researchers.


  • data analysis;
  • mediating variables;
  • mediation analysis;
  • statistics