Can it be fixed without defining initial estimates or respecification?
I entered NLOPTIONS MINITER=## and got down to 2 consecutive integers hoping something in between would allow the model to converge, but nothing worked. "Optimization cannot be completed."
I also tried other optimization methods other than LEVMAR (I was told that you could try many other methods, no guidance on which method is better for specific kinds of data though)
As a possible quick fix if you don't want to alter the model or set starting values, you may want to try using generalized least squares (method= GLS in the Proc Calis statement) instead of maximum likelihood (the SAS default) if you have a decent sample size. GLS and ML are asymptotically equivalent but the GLS fit function is less complex and doesn't require taking the log, so it may be able to converge where ML cannot.
The models often do not covnverge if the research sample is not big enough.
with regrads
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Structural Equation Modeling