Count model with non-estimable estimates

Melk — Thu, 18 Oct 2018 19:50:09 GMT

I am running a count model on number of treatments given and comparing them between 2 diseases. I have excess zeros, but there is no extra "Excess" zeros so I am using a negative binomial model and not a zero inflated model. One of the disease groups is small at about 20 patients. when I run an estimate statement it says "non-est" but I do get MLE parameter estimates. Can I just exponentiate the estimates to get the rate per group although the estimate statement that is doing this is not running?

proc genmod data = dat;
class group;
model numtx = group/ type3 link=log dist=negbin offset=loghospdays;
estimate 'Asthma' group 1 0 / exp;
run;

Re: Count model with non-estimable estimates

PGStats — Fri, 19 Oct 2018 02:00:21 GMT

To get the group mean counts use

lsmeans group / ilink;

To get the ratio of the group mean counts use

lsmeans group / diff exp;

topic Re: Count model with non-estimable estimates in SAS Procedures

Count model with non-estimable estimates

Re: Count model with non-estimable estimates