I'm testing a moderation effect using multiple group analysis in Amos. My question is do I test moderation on all proposed relationships between the indicators and the outcome variables or only on the path with significant relationship? It seems that moderation only works on existing significant relationship between A and B, but I found the non significant relationships also have substantial changes when adding the moderator in (e.g. some non significant relationships become significant). Should I delete the non significant path from the model when testing the moderation? Thanks a lot!
I think the following link may help you. it has different statistical tests for moderation and mediation
Dardas, L., Ahmad, M. (2015). For Fathers Raising Children with Autism, Do Coping Strategies Mediate or Moderate the Relationship between Parenting Stress and Quality of Life? Research in Developmental Disabilities. 36, 620-629. http://dx.doi.org/10.1016/j.ridd.2014.10.047
The following article may help
Dardas, L., Ahmad, M. (2015). For Fathers Raising Children with Autism, Do Coping Strategies Mediate or Moderate the Relationship between Parenting Stress and Quality of Life? Research in Developmental Disabilities. 36, 620-629. http://dx.doi.org/10.1016/j.ridd.2014.10.047
Hi Hailian,
In multiple-group analyses there are two issues wrt, invariance/difference testing
a) there are some parameters that you have to test for equality/invariance in order to compare/test the difference between other parameters. If your interest is to test a moderator hypothesis that the structural effect of latent Variable A on Variable B differs across the groups you *have to* test the equality of factor loadings as a prerequiste test. Here you want to find that the loadings DO NOT differ
b) there are those parameters that you actually have in mind *for theoretical reaons*. That is you expect that some paramateres differ across the groups. In most cases these are the structural effects (e.g., A --> B).
That means
a) test the loadings for equality ("metric invariance")
b) test the parameters OF THEORETICAL INTEREST for equality. You do not have to test every paramater just because it is possible :)
Some are necessary - some are interesting.
See my invariance paper which can be downloaded on my page. In it you will find further relevant references.
Best,
Holger
I forget to add. You said:
>It seems that moderation only works on existing significant relationship between A and B, but I found the non significant relationships also have substantial changes when adding the moderator
Can you clarify this? You have an unrestricted model in which an effect is separately estimated for two groups, ok? And this effect is non-significant in both groups? An when you specify an equality constraint you get a misfit pointing to a difference of both effects? Thats strange but not unlikely, when one of the two effects is just below the singificance "border".
Best,
Holger
Thanks Prof Steinmetz. It was helpful.
For the multiple group analysis, i did the metric invariance test first and then tested the structural path one by one for moderation.
my problem is that in the model, i have two independent variables (let's say A and B) and four dependent variables (e.g. 1, 2, 3, 4 ). for the main test, the result showed that only one independent variable has significant impact on the four dependent variables (A and 1, 2, ,3 ,4). so for the moderation test, a) do i need to test the group difference on the insignificant paths (B and 1, 2, 3, 4 )? b) for the invariance testing (metric), should i test the full model (A, B and 1, 2, 3, 4) or delete the insignificant path and test the model with only significant paths (A and 1, 2, 3, 4 )?
Not sure if i have stated the problems clearly, i will read your papers first and see if i can get some clue.
Regards,
Hailian
Hi Hailan,
I've got the same problem. What did you find out? can we test it in the case of insignificant paths? Please let me know.
Regards,
Sandeep
- Badges
- Science topic
- Similar topics
- Psychology
- Emotional Psychology