10-03-2011 05:51 PM
Hi, I need to analyze count data. My dataset has 20 general practitioners, each of whom has a number of patients screened for kidney stones. So I have the number of patients with stones for each physician as the outcome variable. To account for the fact that each physician has a different total pool of patients and hence a different likelihood of having patients with stones, I have been told that the total number of patients for each physician has to be treated as exposure/offset variable.
My question is: do I need to transform the offset variable? I have read that it is suggested to use the log of the variable, but then again I have only found info for a time exposure variable. Is it the same when the exposure variable is not observation time but rather the total population at risk? Or do I have to use the raw variable instead?
I am using PROC COUNTREG.
Of course thank you for your answers.