How do I tell GENMOD to generate missing values in ESTIMATE statement when warnings are present?

[JamesLin]

Hello:

I am running a macro with a large number of Negative Binomial regressions in GENMOD. I know that in some cases the result is questionable (due to a negative overdispersion parameter) but still GENMOD outputs non-missing results from ESTIMATE in all cases,

including the models that produce warnings like:

WARNING: The relative Hessian convergence criterion of 0.01955056 is greater than the
limit of 0.0001. The convergence is questionable.
WARNING: The procedure is continuing but the validity of the model fit is
questionable.

Is there a way to tell GENMOD to output nothing or missing values when the model fit generates warnings?

[pau13rown · Accepted solution]

this is a common problem when running data simulations and eg bootstrapping from a small sample, some of the samples don't converge. There is an ods 'convergencestatus': https://support.sas.com/documentation/cdl/en/statug/63033/HTML/default/viewer.htm#statug_genmod_sect...

i think this is how i normally do it, ie in the macro catch this variable and later discard estimates from samples that didn't converge. I've seen others do likewise

[JamesLin]

Thanks for replying. I inserted a check based on the Status variable obtained from ods output statement as illustrated here:

https://blogs.sas.com/content/iml/2015/08/26/convergence-sim.html