I conducted two path models with 3 dependent variables and 6 / 7 correlated predictors. I further used multiple imputation and type=complex (due to the hierarchical structure of my data). Although I'm only interested in associations between individual-level-variables, my reviewers want me to control for two second-level-factors (schooltype and federal state). In reaction to this I extended my path models to intercept-as-oucome-models with schooltype and federal state as dummy-coded second-level-factors.
Whereas one of the models shows only minor problems (I get the warning that my standard errors may not be trustworthy because I have more parameters than clusters), I get the following error message for the other model:
THE MODEL ESTIMATION DID NOT TERMINATE NORMALLY DUE TO AN ILL-CONDITIONED FISHER INFORMATION MATRIX. CHANGE YOUR MODEL AND/OR STARTING VALUES.
THE MODEL ESTIMATION DID NOT TERMINATE NORMALLY DUE TO A NON-POSITIVE DEFINITE FISHER INFORMATION MATRIX. THIS MAY BE DUE TO THE STARTING VALUES BUT MAY ALSO BE AN INDICATION OF MODEL NONIDENTIFICATION. THE CONDITION NUMBER IS -0.582D-18.
THE STANDARD ERRORS OF THE MODEL PARAMETER ESTIMATES COULD NOT BE COMPUTED. THIS IS OFTEN DUE TO THE STARTING VALUES BUT MAY ALSO BE AN INDICATION OF MODEL NONIDENTIFICATION. CHANGE YOUR MODEL AND/OR STARTING VALUES. PROBLEM INVOLVING PARAMETER 35.
Parameter 35 is the intercept of one of the second-order-factors on one of the dependent variables. I couldn't find any hint for misspecification and even fixing parameter 35 didn't solve the problem. First I assumed, it could be a problem with MI but I recalculated the model with FIML and came to the same result, except that now parameter 41 is involved. I also already tried to simplify the model to a random-intercept-model but the problem persisted. Do you have any suggestions where the problem might be and how I may get the model to work?