What is Mediation analysis?
Mediation analysis is the statistical procedure for testing whether the effect of one variable on another operates through a proposed intermediate mechanism. It decomposes a total effect into a direct effect and an indirect effect through the mediator.
How it works
The analysis tests a causal chain: X affects M, and M affects Y, therefore X affects Y through M. Modern methods include bootstrapping (which does not assume normal distribution of the indirect effect), structural equation modeling (which handles multiple mediators simultaneously), and causal mediation analysis (which addresses confounding of the mediator-outcome relationship). The key assumption is that the mediator is not confounded with the outcome, which cannot be guaranteed without randomizing the mediator itself.
Applied example
A researcher tests whether a gratitude journaling intervention reduces depression through the mediator of positive emotions. They find a significant indirect effect: journaling increases positive emotions, which in turn decrease depression. This guides clinical practice toward interventions that target positive emotions specifically.
Why it matters
Mediation analysis transforms evaluation from ‘Does it work?’ to ‘How does it work?’, enabling more targeted and efficient intervention design.
