- Frazier, Tix, and Barron (2004, pdf here The article provides a gentle introduction to mediation and moderation along with examples of how to write up such analyses.
- My lecture notes on Mediation
I gave a three hour seminar on the topic with slides and notes
- Kenny's notes on mediation
- Post on mediation and causality
- Links on causal inference in psychology
R Mediation Resources
- mediation package with links to documentation
sobel()function in the
R Moderation Resources
- To examine moderation is to examine an interaction effect.
R makes it very easy to examine interaction effects with
its standard model notation
(see William Revelle's site
for a discussion of
lmand model notation).
moderate.lmin QuantPsyc package
Why I often shudder when I get asked about mediation and moderation in statistical consulting sessions
Researchers testing for mediation (IV to MV to DV) often make causal conclusions when none are warranted. At the very least researchers should consider alternative causal pathways.
Testing for moderation can be sexy in some disciplines in a way that testing simple main effects is not. However, in most observational studies that I encounter in the social sciences, interaction effects tend to be small if present at all. Researchers should not lose site of the importance of main effects in the quest to find supposedly sexy interactions.