Mediation Analysis With Longitudinal Data
David L. Roth
University of Alabama at Birmingham
David P. MacKinnon
Arizona State University
Mediation analysis provides a multivariate framework for testing hypotheses about chains of causal relationships among multiple variables (MacKinnon, 2008). The most straightforward mediation model is one where a primary causal variable X leads to changes on an observed mediating variable M, which in turn causes changes on an outcome variable Y. This relatively simple model is remarkably useful in many research contexts. It can be used, for example, to evaluate the hypothesized causal mechanisms that account for intervention effects on outcomes or to examine the fit of purely observational data to theoretically ...