I am testing an SEM model and want to see if the paths coefficients differ for males and females. I got the following:
UNCONSTRAINED MODEL Value 0.000* Degrees of Freedom 0 P-Value 0.0000 Scaling Correction Factor 1.000 for MLR
Chi-Square Test of Model Fit
Value 4.545 Degrees of Freedom 5
P-Value 0.4739 Scaling Correction Factor 1.152 for MLR
I am confussed by the 0 chi-square and p values for the unconstrained model. If it is a problem, how can I fit it?
I read about using -2loglikelihood in some cases, but I don't know if that applies here given the chi-square values in the unconstrained model or if that is to be used for categorical outcomes. My outcomes are continuous.
I also got the following error. How can I fix this error? And where is parameter 26?
THE STANDARD ERRORS OF THE MODEL PARAMETER ESTIMATES MAY NOT BE TRUSTWORTHY FOR SOME PARAMETERS DUE TO A NON-POSITIVE DEFINITE FIRST-ORDER DERIVATIVE PRODUCT MATRIX. THIS MAY BE DUE TO THE STARTING VALUES BUT MAY ALSO BE AN INDICATION OF MODEL NONIDENTIFICATION. THE CONDITION NUMBER IS -0.467D-17. PROBLEM INVOLVING PARAMETER 26.
Thank you. Will it be accurate to report that value as the chi-square diff test (5) = 4.54, p > .05 ? And then conclude that there are no differences in the structural relations in my model between boys and girls?
What I threw me off was whether the difference test can be computed given the 0 value of the chi-square in the unconstrained model.
Thanks very much! It is awsome that you are available for questions and always giving people feedback.
Ironically enough I have done all my analyses in AMOS to then find out that I need to use MLR in MPlus to answer my moderation question. So far, I've done the analysis with one of the two informants I am using and got the same findings as in AMOS.
I have 2 questions:
Under what circumstances does MPlus give different findings than AMOS?
I looked at the unconstrained model and noticed that one of my paths was significant for boys but not for girls. But then the actual chi-square test is not significant, which leads to the conclusion that the structural relations are the same for boys and girls? Should I ignore that one path for boys was sign but not for girls? In other words, one does not go by the path significance but by what the chi-square diff test says? I remember my professor saying that people confusse the main effects with moderation effects. Will going by the path significance mean falling into that trap?
These are all my questions. Thanks again. I am super excited to now be able to use MPlus.
Mplus and Amos will agree as long as the same data, same model, and same estimator are used.
The chi-square difference test tests whether the coefficients are equal for boys and girls. The z-test tests whether the coefficients are significantly different from zero. I don't think these findings are inconsistent.