I understand that for partial mediation to have occured the CI for the indirect effect must not cross zero and the b value for the indirect effect should be smaller than the b value for the direct effect.
I have a result in which although it is significant (ie CI does not cross zero) the b value for the indirect effect is larger than that for the direct effect. What does this mean?
@Karan Douglas
If the direct effect in your model is significant (be it greater or less than indirect effect) there is partial mediation.
Imran Anwar - please could you point me to any paper(s) or text books that show this. I'm not doubting your answer - but I know my PhD supervisor will ask me! Many thanks
@Karan Douglas
Kindly refer to the website below by David A. Kenny👇🏼 you'll have detailed information.
http://davidakenny.net/cm/mediate.htm#:~:text=Partial%20mediation%20is%20the%20case,outcome%20and%20not%20vice%20versa.
It is almost impossible to answer the question (please pardon me) without knowing the underlying assumptions.
How did you model the mediator? Discrete (0 or 1) or stochastic?
If discrete then the conclusion is simple, but if stochastic, then the modeler needs to assign a passing grade for "significant".
It's hard to answer this question without telling us the significance of all the effects including the direct and total effect. If I read between the lines I think you have partial mediation but I cannot be sure without more information
- Badges
- Science topic
- Similar topics
- Condensed Matter Physics
- Zero