Hi, I am working on cross-cultural research. For that, I need to perform MG analysis in AMOS. I found the following results:
CMIN / DF : 5.619
GFI : .946
AGFI: .934
CFI: .969
RMSEA: .042
For individual country, CMIN value is less than 4 while cumulative is more than 5, while all results are perfectly alright. I think with increasing sample size CMIN value also increase. Need your comments.
Further, should I proceed with the above results with same CMIN value 5.619? Need your expert opinion. Thank you
Hello Asif
The ratio of the chisquare and df is not a very good basis to evaluate the model an the GFI and AGFI are neither. I guess the chisquare test is highly significant?
The N is surely very large but nonetheless you should dig into potential problems of the factor structure. Unfortunately, AMOS does not provide much details that are essential for respecifying a misfitting model (as always I would strongly advise to use R/lavaan, Mplus, or Stata). Perhaps a combination of the inspection of the question wordings (whether it makes sense to perceive them as reflections of the same underlying latent variable) and unusual differences in factor loadings can lead to some kind of idea.
If you present the question wordings and a path diagram, others may have their thoughts, too.
Further, with large samples there is often some substantial amount of population heterogeneity involved. This can be problem, too.
At any rate I would first start to focus on one culture at a time, as probably the basic structure is misspecfied in one of them (or all).
Best,
--Holger
Dear Dr. Holger, thanks for your response. Yes, it is highly significant. My model variables are as follows:
2 IVs 1 Mediator 2 Moderators 1 DV with three reflective constructs
What do you recommend in this situation? Where need rectifications?
Please advise
Thank you
Asif,
I am confused as I recommended a few things (check loadings, question wordings, subroups within culture, post question wordings and model structure). It doesn't help much to know how many IVs etc. you have.
Best,
Holger
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Sampling Design
- Sample Size