Neerav

An excellent 2 paragraph summary.

However, I find Stokes et al. kind of confusing (although it's good bcs it gives lots of SAS code). I prefer Long, JS Regression models for categorical and limited dependent variables Both these cover a lot more than just dichotomous DVs. Another great book on logistic is Hosmer & Lemeshow Applied Logistic Regression.

Also, people should be aware that there are some new things in LOGISTIC, but that some of these have not made there way into the index that get when you get help from within a SAS session, and it's better to look at the chapter The Logistic Procedure in the online SAS-Stat User's gude (e.g. the glogit option only shows up here not in the index

HTH

Peter

>>> Neerav <neerav_monga@CAMH.NET> 11/23/2004 12:06:38 PM >>> Bob,

As Peter said, the odds ratio explains the association between and an IV with a DV. It is as its name suggests, a ratio of 2 odds. So for example, if you're looking at smoking status (yes or no) predicting low birthweight baby (yes or no), with 'no' being the reference category, an odds ratio of 2.0 for example would say: The odds of a smoker having a low birthweight baby is 2.0 times the odds of a non-smoker having a low birthweight baby. This is the categorical case.

As for continuous, it is the same interpretation except you're talking about an odds ratio for a one unit change in your predictor (eg. age). Again , its a ratio of the odds of one event occuring to the odds of another event (smoker vs non-smoker in my example). Hope that helps, if you need more information, you can look at "Categorical data anlaysis using the SAS system", by Stokes, Davis & Koch. A very good SAS book.

Cheers,

Neerav