Hi, I conducted an SEM across two groups with 7 latent exogenous variables (3 to 5 indicators each, except one latent with only 2 indicators) predicting 1 latent endogenous variable with 3 indicators. I ran the model with the overall group; model did not converge in 10,000 iterations. I ran the model separately by groups - in the smaller group (N=465 the model estimation terminated normally and had good fit for that group; in the other group (N=3307) I got the message
THE MODEL ESTIMATION TERMINATED NORMALLY THE STANDARD ERRORS OF THE MODEL PARAMETER ESTIMATES COULD NOT BE COMPUTED. THE MODEL MAY NOT BE IDENTIFIED. CHECK YOUR MODEL. PROBLEM INVOLVING PARAMETER 132. THE CONDITION NUMBER IS 0.178D-12.
From TECH1 output it looks like parameter 132 is the latent endogenous variable in the psi matrix. I'm not sure what the problem is or how to address it. Also,what does the condition number mean (or is there a place I should look for that information). Thanks for your help, Maja
The condition number is the ratio of the smallest to the largest eigenvalue of the estimated information matrix. A very low condition number implies that the information matrix is singular which implies that the model is not identified.
Can we refer to the condition number to identify collinearity? According to Belsley et al. 1980 a condition number of greater than 100 indicates collinearity (i.e the ratio of the largest to smallest eigenvalue). Do you have any cutoff values for mplus in mind?