Second-order CFA among multiple groups
Message/Author
 rebecca tang posted on Thursday, October 14, 2010 - 9:32 am
Hi,

I am working on a project about comparing second-order CFA between two groups.

Suppose there are four constructs x, y,z, m
x has several measurement scales, x1, x2,x 3,....
y has several measurement scales, y1, y2, y3.....
z has several measurement scales, z1, z2, z3,....
I also link m (no measurement scale)--> x, m-> y, m->z.

Because it is the second-order CFA, among three paths m->x,m->y, m->z, one of them has to be restricted to 1. When I compare the second-order CFA, how can I compare the path that is already restricted to 1? Suppose the path m->x is restricted to 1. After I compare the path m->y, m->z, can I free the path m->z, but restrict m->y as 1. And then I compare the path m->x. Is it possible?

Thank you very much for your help. It is really helpful for me!
 Linda K. Muthen posted on Friday, October 15, 2010 - 9:10 am
Free all of the factor loadings for the second-order factor and set its metric by fixing the factor variance to one.
 songthip ounpraseuth posted on Tuesday, February 01, 2011 - 12:14 pm
Hi Dr. Muthen –

I wanted to examine a second order CFA component prior to testing my larger SEM model. One of my construct is given below:

USEVARIABLES ARE X1-X12;
CATEGORICAL ARE X1-X12;
MISSING ARE ALL (-99);
MODEL: F1 BY X1 X2;
F2 BY X3 X4;
F3 BY X5-X8;
F4 BY X9-X12;
F5 BY F1 F2 F3 F4;

OUTPUT: Standardized; modindices;

It’s failing to provide estimates since the number of iterations exceeded and no convergence was achieved.

Is the problem caused by the fact that F1 and F2 have only two first-order factor indicators? A colleague ran the same model in AMOS which provided various estimates and fit statistics. Although the output appeared normal, I was concerned that the estimator used was MLE while in the presence of ordinal indictors.

I am currently running Mplus version 6.0.

Any insight or suggestions would be much appreciated.
 Linda K. Muthen posted on Tuesday, February 01, 2011 - 12:53 pm
Please send the full output and your license number to support@statmodel.com so I can see why the model did not converge.