I estimated a log-rate model. I had a sample of individuals from
different cities, and the number of respondents in city A differed from
the number of respondents in city B, and so on. I estimated a model where
I used the offset to control for the varying exposure.
I sent the paper off, and a reviewer said I needed to control for city
size. I am not sure I should do this. I could see an argument both ways.
One could argue that my control for varying exposure addresses the issue
of size. One could also argue that the varying exposure control only
addresses variation in sample size per city, and that to really address
the different experiences of being in larger or smaller cities I should
control for city size.
Thanks for any comments you can provide.
Sam
|