topic Re: proc genmod from sas 9.2 to sas 9.4 in Statistical Procedures

proc genmod from sas 9.2 to sas 9.4

Chiara1 — Tue, 24 Jan 2017 16:44:05 GMT

Dear all,

I have a problem with PROC GENMOD, exactly the passage from sas 9.2 to sas 9.4.

The code is:

proc genmod data = datasetname1;
class GROUPN factor2 factor3;
model var_count = GROUPN factor2 factor3 covariate1/ TYPE3 offset=logtime dist=negbin;
lsmeans GROUPN / ilink cl;
estimate 'TRT1 vs TRT2' GROUPN 1 -1 / exp;
run;quit;

I added ILINK option because in this way the results from sas 9.2 and the results from sas 9.4 will be the same.

BUT sometimes, in sas 9.4 the model not converge (while in sas 9.2 converge) and the results are not the same. I try to add iterations with MAXITER=200 and the model in sas 9.4 converge BUT the results still not the same as the results in SAS 9.2.

I think there is a defoult option chenges from sas 9.2 to sas 9.4 but I don't understand wuich.

Can you help me?

Thank you

Cheers,

Chiara1

Re: proc genmod from sas 9.2 to sas 9.4

Rick_SAS — Tue, 24 Jan 2017 18:31:41 GMT

I suggest you look at this thread

https://communities.sas.com/t5/SAS-Statistical-Procedures/Question-re-Consistency-of-GLIMMIX-results-between-9-3-and-9-4/m-p/166465#M8700

which is similar. To summarize the discussion there: Many possible changes can lead to different behavior, including compiler changes, 32-bit vs 64-bit versions of SAS, changes to the underlying algorithm, etc. These problems tend to be compounded if you specify a model that does not fit the data well.

Re: proc genmod from sas 9.2 to sas 9.4

Chiara1 — Tue, 02 May 2017 12:14:14 GMT

With the addition of NOLOGNB option (added with version 9.4), the results will be the same as in version 9.2

Regards