Interaction and moderation quite similar but slightly different:
- Interaction (more general) = effect of X on Y different at different values of Z
- Moderation (more specific) = moderator Z is predictor that changes effect of X on Y
Baron & Kenny method to test mediation
Problems with baron & Kenny method:
- Statistical significance of indirect effect is not formally tested
- Different parts tested separately type I error (false positive)
- Low statistical power
Mediation test with: Sobel test (problematic when not normally distributed) or
bootstrapping
Result meta analysis in forest plot
When <1.0 favours treatment
When >1.0 favours control
Diamond is overall effect size
- Interaction (more general) = effect of X on Y different at different values of Z
- Moderation (more specific) = moderator Z is predictor that changes effect of X on Y
Baron & Kenny method to test mediation
Problems with baron & Kenny method:
- Statistical significance of indirect effect is not formally tested
- Different parts tested separately type I error (false positive)
- Low statistical power
Mediation test with: Sobel test (problematic when not normally distributed) or
bootstrapping
Result meta analysis in forest plot
When <1.0 favours treatment
When >1.0 favours control
Diamond is overall effect size
