The literature seems to tie this to low statistical power (adding M increases explanatory power of the model?), but my hunch is that this could occur due to one or more omitted variables and/or an underlying recursive model, otherwise I cannot wrap my head around it.
Dear Metin Bulus
It depends, Even if there is no overall effect, equate mediation by the existence of an indirect effect. The significance of total effect is not required for claiming mediating effect (indirect effect).
Best Regards
Dr. Fatemeh Khozaei
Metin Bulus
This could happen due to a suppressor effect. A suppressor variable S, when added to the model, can "enhance" the ability of an independent variable X to predict Y by suppressing irrelevant variance in X. See, for exampleArticle MacKinnon DP, Krull JL, Lockwood CM. Equivalence of the medi...
This can happen if the direct and indirect effect have opposite signs, so that they cancel each other out.
Dear Metin Bulus
while understanding a mediating effect, try to think from a theoretical perspective rather than a statistical one (such as of Baron and Kenny, 1986), because inferences we draw from a scientific study are products of theoretical logic and not the mathematics of the statistical procedures we employ to understand our data. I think it will help you "wrap your head around" it :)In fact there are always omitted variables in a study and that is justifiable through Scope/Limitations of the study. Tq.
This is the case of full mediation. As the the relationship between X and Y is fully mediated by the mediator M
- Badges
- Science topic
- Similar topics
- Psychology