What Is the PROCESS Macro? Testing Mediation, Moderation, and Model Numbers
Daniel Ha · Seoul National University PhD StudentWhat Is the PROCESS Macro?
The PROCESS macro, developed by Andrew Hayes, is an extension for SPSS and R that performs regression-based mediation and moderation analysis. Instead of running several regressions by hand, you specify a model number and the indirect effects are computed automatically.
What Does PROCESS Test?
It tests mediation (why and how), moderation (when and for whom), and the combination of the two, moderated mediation.
| Effect | Question |
|---|---|
| Mediation | Does X affect Y through M |
| Moderation | Does the X to Y relationship change with the moderator |
| Moderated mediation | Does the mediation effect itself change with the moderator |
How Do You Use the PROCESS Macro?
Work in order from installation to interpretation.
Step 1: Install and Choose a Model
Install PROCESS in SPSS and select the model number that matches your analytic structure.
Step 2: Assign Variables
Assign the independent variable (X), dependent variable (Y), mediator (M), and moderator (W).
Step 3: Configure Bootstrapping
Set the number of bootstrap samples (for example 5,000) and the confidence level used to estimate the indirect effect.
Step 4: Interpret the Results
Check the bootstrap confidence interval for the indirect effect, and judge moderation from the significance of the interaction term.
How Do You Judge a Mediation Effect?
If the bootstrap confidence interval for the indirect effect does not include zero, the mediation effect is considered significant. Bootstrapping is recommended over the older Baron and Kenny causal steps approach because it is more accurate.
What Are the Model Numbers?
They are the numbers Hayes assigned to each mediation and moderation structure. Simple mediation is Model 4, simple moderation is Model 1, and moderated mediation is Model 7 or Model 14, among others. Choose the number that matches your conceptual model.
Summary
The PROCESS macro tests mediation and moderation through model selection, variable assignment, bootstrapping, and confidence interval interpretation. Judge indirect effects by whether the confidence interval excludes zero, and pick the model number that fits your conceptual model. For complex models with latent variables see structural equation modeling, and for basic effect estimation see regression analysis.