st: Dummy Variables vs. Subgroup Models in Logistic Regression

[brian.h.nathanson@att.net]

Dear Stata Users,

     I'm creating a logistic regression model with many dichotomous variables along with one term that has 8 categories coded 1,2,..8.  I can create 7 dummy variables and have a very large model.  Would it be legitimate if my sample sizes are large enough to create 8 separate models with each model representing one subgroup?   Can anyone comment on the pros and cons of using dummy variables versus creating separate "subgroup" models based on the remaining independent variables?  Thanks!

   -Brian