Moderated Mediation
Message/Author
 Estee posted on Wednesday, November 13, 2013 - 8:48 am
Hi,

Good day.

1)I have tested a model with 2 independent variables (IV), 3 mediators (M) and 3 dependent variables (DV). Now, I am interested to test whether gender (1 = Male 2 = Female) moderates those relationships from IVs to mediators and from mediators to DVs. Shall I use multi group approach? If yes, is there any reference for the syntax?

2)I also read the paper: "Preacher, K.J., Rucker, D.D. and Hayes, A.F. (2007). Addressing moderated mediation hypotheses: Theory, methods, and prescriptions. Multivariate Behavioral Research, 42(1), 185-227."
I found that the moderation effect that i want to test is called "moderated mediation", which is correspond to Model 5 showed in the paper. I found the example of the syntax, yet, the example contain only 1 IV, 1 mediator and 1 DV. I wonder how's the syntax should be if in my case, there are 2 IVs, 3 mediators and 3 DVs?

3) Between multi group approach and Preacher's, which approach is more recommended?

Thank you very much..
 Linda K. Muthen posted on Thursday, November 14, 2013 - 8:39 am
With a binary moderator, it is more straightforward to use the multiple group approach. I would use that.
 Estee posted on Thursday, November 14, 2013 - 7:14 pm
Alright, thank you. If i use multigroup approach, let say the syntax is like this

!Direct relationship from IV to DV
A on C (P1);

!Relationship from IV to mediator
A on B (P2);

!Relationship from mediator to DV
B on C (P3)

For model Female, what's the syntax?

Thank you.
 Linda K. Muthen posted on Friday, November 15, 2013 - 11:48 am
Compare

MODEL:
a ON c;
a ON b;
b ON c;

to

MODEL:
a ON c (1);
a ON b (2);
b ON c (3);
 Estee posted on Saturday, November 16, 2013 - 12:06 am
Thank you very much. I understand now.
 Estee posted on Saturday, November 16, 2013 - 3:04 am
Hi,

I ran a a model with 2 IVs, 3 mediators, 3 DVs. There are totally 21 paths for all direct and indirect relationships. As mentioned previously, I want to see whether gender (male/female) moderates these paths.
First, I test a fully unconstrained model. Then, a fully constrained model. The chi square difference is significant. Subsequently, I constrain each of the paths one at a time to see which particular paths are moderated by gender. Out of all 21 paths, there is 1 path which the result shows the error " NO CONVERGENCE. NUMBER OF ITERATIONS EXCEEDED."

I read the user guide about the problem of no convergence. But I am not sure about the solution.