A health psychologist tests whether caffeine consumption moderates the relationship between stress and sleep quality (higher scores = better sleep) in a sample of N = 200. Stress and caffeine were mean-centered before creating the interaction term.
SPSS Hierarchical Regression Output (DV = Sleep Quality)
Model Summary
- Model 1 (Stress_c, Caffeine_c): R² = .203
- Model 2 (+ Stress_c × Caffeine_c): R² = .240
- ΔR² = .037, F-change(1,196) = 5.92, p = .016
Coefficients (Model 2)
- Stress_c: b = -0.40, SE = 0.10, p < .001
- Caffeine_c: b = -0.05, SE = 0.09, p = .580
- Stress_c × Caffeine_c: b = -0.28, SE = 0.12, p = .016
Which interpretation is MOST correct?
There is evidence of moderation: the negative effect of stress on sleep quality becomes stronger (more negative) as caffeine increases; because predictors are centered, bStress is the effect of stress when caffeine is at its mean, and the next step is to probe/simple-slopes (e.g., ±1 SD caffeine).
There is no moderation because caffeine’s main effect is not significant; therefore, the interaction should not be interpreted and Model 1 is the final model.
This output suggests mediation: caffeine explains why stress affects sleep quality, because the interaction term is significant and ΔR² increased.
There is evidence of moderation: the negative effect of stress on sleep quality becomes weaker (less negative) as caffeine increases; since predictors are centered, the constant is meaningless and should be ignored.