Moderated Mediation Analysis
Also known as: conditional process analysis, moderated mediation model, first-stage moderated mediation, second-stage moderated mediation
Moderated mediation tests whether the indirect effect of an independent variable on an outcome — transmitted through a mediator — differs in strength depending on the level of a moderator variable. It answers the question: for whom, or under what conditions, does the mediated pathway operate most strongly?
Read the full method
Sign in with a free account to read this section.
Method map
The neighbourhood of related methods — select a node to explore.
When to use it
Use moderated mediation when theory specifies both a mechanism (how X affects Y) and a boundary condition (for whom or when that mechanism operates). The sample should be large enough to detect interaction effects reliably; n of at least 200 is a common guideline, because interactions in regression require more power than main effects. All variables should be measured validly and reliably, as measurement error attenuates both mediated and moderated paths. Do not use moderated mediation when you lack a theoretically motivated moderator, when the mediator is suspected to be endogenous (prefer SEM with latent variables), or when the mediation pathway itself has not been established in prior work and exploratory moderation of an uncertain mechanism would be speculative.
Strengths & limitations
- Integrates mechanism and boundary-condition hypotheses in a single, coherent model rather than running separate mediation and moderation analyses.
- The bootstrap confidence interval for the conditional indirect effect is valid regardless of the shape of the sampling distribution of the product term.
- Hayes's PROCESS macro implements the full range of moderated mediation templates in SPSS and R, making the method accessible without bespoke programming.
- Provides an index of moderated mediation that formally tests whether the indirect effect changes across the range of the moderator.
- Applies to a variety of outcome types through generalized linear model extensions.
- Interactions in regression require substantially larger samples than main effects, and underpowered studies produce unstable conditional indirect effect estimates.
- The method uses observed (error-laden) variables; measurement error in the mediator or moderator attenuates the interaction term and biases the conditional indirect effect.
- Many published templates assume linearity; non-linear moderation of the indirect effect requires more complex specifications that are rarely applied.
- Causal interpretation requires that X precede M, M precede Y, and the model be correctly specified — assumptions that cannot be verified from observational cross-sectional data alone.
Frequently asked
What distinguishes moderated mediation from mediated moderation?
In moderated mediation the primary research question concerns the indirect effect: does the X-to-Y mechanism via M depend on W? In mediated moderation the primary question concerns the interaction: does a previously established X-by-W interaction on Y operate through M? In practice both can be estimated from the same model, and Hayes (2018) argues the conditional indirect effect is the most interpretable quantity in either case.
Do I need to center my variables before running the model?
Mean-centering X and W before computing their product term is strongly recommended. It makes the lower-order coefficients (a1, a2, c-prime) interpretable as effects at the mean of the other variable, and it reduces the numerical collinearity between the interaction term and its components. The conditional indirect effect itself is unaffected by centering.
How many bootstrap samples should I use?
A minimum of 5,000 is recommended for 95% confidence intervals; 10,000 or more is preferred for 99% intervals or when reporting the index of moderated mediation. Using too few resamples can produce unstable interval boundaries.
Can moderated mediation be estimated with latent variables?
Yes. SEM-based approaches such as the latent moderated structural equations (LMS) method or Bayesian product-indicator methods estimate moderated mediation with latent variables, correcting for measurement error. These are more complex and require specialized software (Mplus, lavaan with extensions) but are preferred when scale reliabilities are modest.
What is the index of moderated mediation?
It is the partial derivative of the conditional indirect effect with respect to W — in the simple first-stage model, this equals a3 times b1. A bootstrap confidence interval for this index that excludes zero constitutes formal evidence that the indirect effect changes as a function of W, which is the defining feature of moderated mediation.
Sources
- Hayes, A. F. (2018). Introduction to Mediation, Moderation, and Conditional Process Analysis: A Regression-Based Approach (2nd ed.). Guilford Press. ISBN: 978-1462534654
- Preacher, K. J., Rucker, D. D., & Hayes, A. F. (2007). Addressing moderated mediation hypotheses: Theory, methods, and prescriptions. Multivariate Behavioral Research, 42(1), 185–227. DOI: 10.1080/00273170701341316 ↗
How to cite this page
ScholarGate. (2026, June 3). Moderated Mediation Analysis. ScholarGate. https://scholargate.app/en/statistics/moderated-mediation
Which method?
Set this method beside its closest kin and read them side by side — the library lays the books on the table; the choice is yours.
- Mediation AnalysisStatistics↔ compare
- Moderation AnalysisCausal inference↔ compare
- Path AnalysisStatistics↔ compare
- Structural Equation ModelingResearch Statistics↔ compare