Exact Poisson, proc genmod

[CrystalS1]

I am trying to compare the incidence rates among 3 groups using exact Poisson regression(the sample size is small, and events are rare). My code is as below:

proc genmod data = data; class B/param=glm ref=first; model A = B/dist=poisson link=log offset=log_PY ; exact B; run;

where A is the event indicator, B is a 3-level categorical variable, and log_PY is log(person-years).

I got a message "Note: # indicates that the conditional distribution is degenerate.", and the exact estimates were not produced. The Exact Conditional Tests looks like below:

             Exact Conditional Tests 
Effect   Test   Statistic   p-Value  Exact Mid 

 B       Score          . #       .          . 

    Probability    1.0000 #  1.0000     0.5000 

I am new to this and wonder if anyone could help me to understand why I got this kind of outputs and how to solve the problem?

If any other SAS proc or R package can do the job, please let me know.

Thanks a lot!

[StatDave]

Try changing PARAM=GLM to PARAM=REF.