Hello Fajle,
The answer depends on what model you're evaluating (as well as which coefficient and which effect you're referring to: unfortunately, neither is clear from your query).
In the simplest mediation model: A -> M -> Y, the correlation, r(M,Y) gives the standardized direct and total effect of the mediator, M, on Y (there is no indirect effect of the mediator in this model). Similarly, the correlation r(A,M) gives the direct and total effect of A on M.
There is an indirect effect of A on Y, which path rules tell us is the product of the A to M path coefficient multiplied by the M to Y path coefficient. In this model, that is the "total" effect of A on Y.
However, if the model changes so that you also allow a direct path from A to Y, then neither the direct nor the indirect effect of A on Y is necessarily equal to the total (unless one of the effects is zero). As well, if M influences some other variable (say, "X"), which in turn influences Y, then no single coefficient can capture the direct or total effect of M on Y.
Good luck with your work.
No the coefficient for the total effect is the one where you exclude the mediator, say c'.
The coefficients when you include the mediator are the indirect effect, a*b, and the direct effect, C where a is impact of X on M and b the impact of M on Y in that model..
The two regressions are comparable such that
c' = c + ab.
That is the total effect from your first regression is the sum of the direct effect and indirect effect from the second model.
Hope that helps.
Check out following for some useful diagrams
https://journals.sagepub.com/doi/full/10.1177/0022243719881618
- Badges
- Science topic
- Similar topics
- Social Psychology
- Attitude