Since I don't have much experience in research, I need your help. Is correlation a requirement in mediation analyses? In my study, the relationship between DV and IV was not significant. However, when I put gender and age as control variables, this relationship turns out to be significant. So in short, there is a partial correlation between DV and IV. Can I still continue with mediation analysis? Should I share data on partial correlation in the article?
From what I've looked into, Hayes says correlation is not a requirement. How can I explain this situation to the jury where I will defend my thesis?
I use PROCESS Macro in my analysis. Significant effects were found for X on Y and M, and for M on Y.
X=IV
Y=DV
M=Mediator
A direct correlation between the independent variable (IV) and dependent variable (DV) is not an absolute requirement to proceed with mediation analysis. Mediation can exist even in the absence of a significant total effect between the IV and DV. This is termed "indirect-only mediation".
However, it is important that your hypothesized mediator is significantly correlated with both the IV and the DV in the mediation model. So examine these paths.
The fact that controlling for age and gender caused the IV to DV path to become significant suggests that those variables influence the relationship as covariates. This would mean you have found a significant partial correlation between the IV and DV while controlling for those variables.
In your write-up, you should report that the total/direct effect of IV on DV was non-significant. But when controlling for gender and age, a significant partial correlation between the IV and DV emerged. You can interpret what this more complex relationship entails.
Then proceed to test the full mediation model and explain the nature of the indirect effects via the mediating variable and its specific effects when controlling for age and gender.
- Badges
- Science topic
- Similar topics
- Philosophy
- Esthetics
- Interpretation