CML reports errors, but return code at end reports success?


Dear GAUSS Users,

I have a question about the return code in CML. I'd like to know whether there is a known issue with it. I am running a regression and want to store the return code in a vector. However that vector is only zeros, which would indicate normal convergence, but there are error messages present in my simulation. I ran another regression with a CO (constrained optimisation) algorithm, and identical storing setup, and there it works.

I hope somebody can help me,



Tags: asked July 25, 2013

2 Answers


There are many types of errors CML may encounter during the iterations from which it can recover and eventually produce a normal convergence.  These errors may give some indication about your model or data, but the normal convergence indicates that it still arrived at the correct estimate.


Dear Ron,

yes that was indeed the case, I changed the maximum amount of iterations to two, and now I get returncode = 2 in every case.

thaks a lot!