Moderation analysis
Ask when or for whom a relationship differs
Moderation means that a relationship changes with a third variable and is usually represented by an interaction term. Simple slopes, predicted contrasts or Johnson-Neyman regions describe conditional effects. Moderated mediation additionally asks whether an indirect effect varies with a moderator. PROCESS automates common models, but coding, model specification and uncertainty still need justification.
Use when theory predicts effect heterogeneity across contexts or people. Plan the interaction, scale and moderator range in advance, with sufficient precision to distinguish meaningful conditional differences.
Strengths
- Tests heterogeneity directly rather than comparing separate p-values
- Produces interpretable conditional predictions
Limitations
- Interactions often require more information than main effects
- Sparse moderator regions encourage unstable extrapolation
Know the boundary
Significant in one subgroup and nonsignificant in another is not itself evidence of moderation.