Message/Author |
|
|
Hi, I'm running a mediation analysis and I have run separate simple mediation models to test the impact of 2 mediators on the relationship between an IV and DV. I now want to test which is the better mediator by running a multiple mediation model containing IV M1 M2 & DV. I have read the paper by preacher & Hayes (2008) who recommend bootstrapping to do this. How do I interpret the output for which mediator best accounts for the relationship between IV & DV? Output: (std) total direct effect =.737, Mediator 1: indirect=.219; CIs [.022,.211](+/-95%) Mediator 2: indirect=.117;[.049,.390]. How would you suggest interpreting this and is this okay to do with a relatively small N? Can I conclude that mediator 1 accounts for more of the relationship? |
|
|
If the results come from a run with both mediators in it, it looks like you need both. If you leave out one of them, the effects for the other are distorted. Mplus does not have a test of one effect being significantly better than the other, only a test of equality. |
|
|
I've run each mediator in a simple mediation model (which were sig for partial mediation) and I've run them as simple models controlling for the other mediator. With mediator 1 and mediator 2 controlled for the model is sig, with mediator 2 as a mediator and mediator 1 controlled the sig mediation is lost. Would the best way to account for the more important mediator to refer to this or the overall R square figure for the simple model looking at which accounts for more variance? |
|
|
You may want to take this general model question by SEMNET. |
|
|
Ok thank you for the advice! |
|
Back to top |