I am performing a mediation analysis using PROCESS by Andrew Hayes, and I don't understand what this output means. I want to see whether rumination mediates the relationship between co-rumination (X) and relationship satisfaction (Y), however I don't understand the indirect effect. Path B is negative, and the indirect effect is negative whereas the direct effect is positive. How to interpret this? Path A= .12, p=.03 Path B = -.61, p=.0001 Direct effect = b=.41, p=.002 enter image description here


"How to interpret this? Path A= .12, p=.03 Path B = -.61, p=.0001 Direct effect = b=.41, p=.002"

a*b is negative and it is (-.0732); c' path (direct effect) is 0.41. And all are significantly different from zero. As you can see the direct effect and the indirect effect have two different sign. This is a typical case of inconsistent mediation (MacKinnon, Fairchild, and Fritz (2007). This has been well addressed in other threads very much! you might be benefited looking into it.

The most important thing to notice is that, in case of suppression effect (another name for inconsistent mediation); the only requirement is the direct effect to be larger in magnitude than the indirect effect. This is clear in your output and you can claim inconsistent mediation. So it means without using jargons "An increase in a decreases b which in turn increases c". You decide if this makes sense, as you know what each variable means and I don't. Please note that in case of inconsistent mediation ab/c does not make sense, so do not report it.


According to my knowledge, path "a" says what is the expected change in your mediator variable, when your independent variable increases by 1 unit. So in your case, it would be that increase in co-rumination (X), increases rumination (M).

Path "b" says what is the expected change in your dependent variable when mediator increases by 1 unit (controlling for X). So in your case, it would be that increase in rumination (M) decreases satisfaction (Y).

Then, it makes sense that indirect effect is negative when you multiply a*b (positive time negative values gives the negative). Looking at your output, it seems that the indirect effect is significant because the confidence interval does not include 0.

This is, to me, a bit strange that you have a positive relationship between co-rumination and satisfaction but I know nothing about the theory.

Maybe this paper would help you understand mediation more https://www.sciencedirect.com/science/article/pii/S0005796716301887


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.