regression
Moderation analysis
Test whether one numeric predictor changes the association of another with an outcome.
Requirements
- numeric outcome
- numeric focal predictor
- numeric moderator
Avoid when
- interaction form is implausible
- causal interpretation without a defensible design
Formula
y = β₀ + β₁x + β₂m + β₃xm + ε
Interpretation
The interaction coefficient quantifies how the focal predictor slope changes across the moderator.
Worked example
Test whether baseline severity changes the association between treatment dose and outcome.
Open example datasets