Peter & David, Thank you for your remarks, I'm glad I was able to help fellow SAS users out there. I am a biostatistician (Msc.) from University of Toronto.

Peter: I agree that Koch's book is weak on interpretation and I realized that you can use proc logistic for polytomous logistic regression (even though they claim you can't specify a link function=glogit, which is untrue as I have done this). Hosmer & Lemeshow is a classic, I have the 1st edition (1989) but might be a bit heavy on the theory side for some, but a great book nonetheless.

I use the online doc very frequently as it does contain a lot of syntax on what options can be implemented. Something that isn't obvoius is that sas 8.2 does in fact do exact logistic regression, although the online doc for 8.2 doesn't mention it, you can find information on the syntax and details in the SAS online doc verion 9 for some odd reason.

David: I agree, the units cannot be lost and it is nice that proc logistic has a UNIT statement that allows us to specify how many units we want the odds ratio in (eg. for 10 years instead of 1 year in my example).

Bob: Happy to help, feel free to ask any follow up questions you may have.

Cheers,

Neerav