Hello Everybody
I have tested the direct relationship between my IV =>M and between M =>DV using linear regression and the relationships were significant. Nevertheless, when I try the same variables with the same control variable on the macro process (Hayes), the relationship between IV and M becomes insignificant further the total indirect effect is not significant also. Is it possible? Any help with references would be highly appreciated.
Hello Khaled,
yes of course. That's why you have to perform a separate statistical test of the indirect effect (e.g., via bootstrapping). The indirect effect follows by the product of both direct effects, hence, it will be much smaller than both direct effects leading to the question about the necessary signal-to-noise ratio in order to conclude that their is indeed an indirect effect in the population.
Best,
--Holger
Additionally to Holger Steinmetz answer: You wrote "... and between M =>DV using linear regression ...", which sounds as if you tested a bivariate regression between M and the DV. But this is not important for the indirect effect, it is the relationship between M and DV (often denoted b), when IV is also in the model. Maybe you did this in the first place, but it was not clear from your description. As Holger mentioned the indirect effect is the product of b times the regression weight of IV on M (a), so a*b.
Thank you for your explanation. In the model it is hypothesized that a direct relationship exist between IV and M, M and DV and then we hypothesized that M is a mediator. That's why I have tested the direct effects before testing the mediation.
- Badges
- Science topic
- Similar topics
- Statistics
- Survey