When there is an error in the analysis, Mplus often reports the number of the parameter that is causing the problem (ex: "PROBLEM INVOLVING PARAMETER 28"). How do I know which parameter corresponds to the number?
If you get a non-identification message, you should get TECH1 automatically if you are using the most recent version of Mplus. If you don't, you should send the output and your license number to email@example.com.
Go to the beginning of TECH1 and use the Find function to find the parameter number that you want. The remedy depends on the situation. If it is not clear to you, send your output and license number to firstname.lastname@example.org.
Is it possible to determine if a parameter is identified without using MODINDICES?
We had been advised to use the following steps when encountering the following error:
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 0.458D-12. PROBLEM INVOLVING PARAMETER 23.
1. Re-run the model with OUTPUT: SVALUES; 2. Use the SVALUES in the resulting output in a new run of the model. a. Set the problem parameter to zero. b. Set STARTS equal to zero. c. Request OUTPUT: MODINDICES (ALL); 3. Check the modification index for the fixed parameter. If it is different from zero, the parameter is identified and the message can be ignored. If the modification index is equal to zero, the parameter is not identified and must be fixed to zero in subsequent runs.
Unfortunately, I just tried that for an LTA and got the following error:
*** WARNING in OUTPUT command MODINDICES option is not available for TYPE=MIXTURE with more than one categorical latent variable. Request for MODINDICES is ignored.