03-13-2012 01:12 PM
I have a dataset with a count variable which takes the values 0,1,2 and three other columns which include smoking, alcohol and bmi information (they are all categorical variables). I also have an exposure column which I am planning to use as an offset variable in my poisson model.
I was wondering whether someone provide me with some information on how to write a code that will enable me to decide which is the best fitted model using the BIC criterion? I would like the code to look among all possible models in this case 8 models. I am planning to extend the code later to include a large number of variables.
I really hope you can help.
03-13-2012 01:32 PM
If you are using SAS version 9.3, look as experimental procedure TCOUNTREG. The statement MODEL ... / SELECT=(INFO(CRITER=AIC or SBC)) gives you model selection capabilities based on information criteria.