Hello All
My question pertains to a strange situation in mediation analysis. Let us suppose my 1st hypothesis says A affects C. Then, my 2nd hypothesis says B mediates relationship of A and C.
If I find support for indirect effect for B (B indeed mediates), what happens to my 1st hypothesis? How do I interpret it? If the support is not found for 1st hypothesis then how do you define this case?
I have seen papers where they have merely reported that A does not cause C (if 1st hypothesis isnt supported) and then state that B mediates A--->C relationship (if 2nd hypothesis is supported).
Am conceptually not able to make sense of this. Does this mean that while A does not affect C (Since the relationship is nonsig.), it does affect C through B? Isnt this contradictory?
PS: I am talking about complimentary mediation here, where paths A to B, and B to C have the same sign coefficient.
Besides checking the A-C relationship directly, you should also examine the direct effect while including the mediation in your model (this effect is often referred to as c'). If c' < c (normal direct effect), then you have evidence of a (partial) mediation.
Theoretically, you can make sense of this by conceptualizing an opposite force for the effect of A-C, meaning that besides mediator B, you theorize that some other mechanism (say D), has the opposite effect. Then, when you look at A-C, both effects materialize and you find no effect. However, when looking at A-B-C (or A-D-C), you will find a mediation effect.
Hello Robert
Wonderful answer and it makes total sense, thanks for that. Basically, what you are saying is the relationship of A-C is getting reduced in size due to two conflationary mediations (A-B-C is positive and A-D-C is negative), which cancel out the total effect.
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Structural Equation Modeling