Calculating crude estimate with proc genmod poisson regression

[sasgorilla]

Hi. I'm starting to do a poisson regression with proc genmod and wanted to first check that I have my data summarized correctly. I assumed that I could test this with the following crude model without covariates and compare the estimate with a by-hand calculation of outcome/person-time. 

proc genmod data=summarized;
	MODEL outcome=  /dist=poisson link=log offset=logdenomcount;
	estimate "Crude" intercept 1;
run;

However,  the estimate and my by-hand rate are not the same. Is my assumption these should match incorrect? 

[sasgorilla · Accepted solution]

Ah, I just went back and realized I made my offset off of the _Freq_ when I did my proc summary and not the summed denominator count. Therefore my numbers were off. 

Correcting that now produces a match with my by-hand calculation as I thought should be the case. 

[Ksharp]

If you want to get RATE ,check "Mean Estimate" column .

If you want to get COUNT, check "L'Beta Estimate" column.