I'm running a GMM specifying 2-4 classes on a continuous outcome for 4 time points. The 2 class model runs without a problem, but the 3 class model gives me a warning that the latent variable covariance matrix is not positive definite. Based on a preliminary LGC analysis including a quadratic slope, I specified the slope coefficients as fixed in the GMM. What is interesting is that when I run a 4 class model, it runs without a problem (i.e., normal convergence and no warnings). Do you have any suggestions for why this might happen for a 3 class model, but not a 4 class model? Are the fit indices (e.g., BIC) and the test for k-1 classes trustworthy when you get a warning like this?
If you get such a warning, the results should not be interpreted. If you send your input, data, output, and license number to firstname.lastname@example.org, I can take a look at it.
Yoon Oh posted on Monday, April 29, 2019 - 1:32 pm
I was running Growth Mixture Models with up to 7 classes. All models with 2-4 classes and 6-7 classes ran without any problem, but only the model with 5 classes encountered the following warning message. Would you please help me how to address this problem? Thank you. -- Yoon
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 0.754D-11. PROBLEM INVOLVING THE FOLLOWING PARAMETER: Parameter 20, %G#4%: [ Q ]