Was wondering if any of you had encountered the following problem:
I am running a simple three factor CFA with second order factor. As my sampling scheme is complex sampling, I am using TYPE = COMPLEX. This is fine until I also wanted to validate the model with another sample, and decided to run a multigroup CFA (TYPE = COMPLEX MGROUP). In this case, Mplus gave the following message:
"THE STANDARD ERRORS OF THE MODEL PARAMETER ESTIMATES COULD NOT BE COMPUTED. THE MODEL MAY NOT BE IDENTIFIED. CHECK YOUR MODEL. PROBLEM INVOLVING PARAMETER 59."
I can resolve this by constraining all the second order loadings (there were only three for my model) or some loadings and the latent factor variance of the seconder order factor. But this doesn't seem like the theoretically correct thing to do.
Further testing showed that even if I did not use TYPE = COMPLEX, but use TYPE = MGROUP; ESTIMATOR = MLR, I get the same problem.
So my question is: Is this really an identification problem with MLR, and is constraining the loadings the correct solution?