Divide population according to predefined predicted probabilities cut-offs

[Panagiotis Antiochos]

Dear all,


I run a logistic regression model with predictors (age, diabetes, hypertension, tobacco) and a dependent variable/outcome (death).


After performing the logistic regression, how can i divide my population according to predicted probability of the outcome? I want for example to use probability cut-offs of 4%, 15% for the outcome and divide my population in three parts (low risk <=4%, intermediate risk 4< <=15%, high risk >15%) of predicted probability for the outcome.

Is there an easy way to do that in STATA?

I appreciate very much any help.
Best,


*************************************************

Panagiotis Antiochos, M.D.

Post-doctoral Research Fellow

Cardiac Magnetic Resonance Laboratory

Brigham and Women's Hospital

75 Francis Street, Boston, MA 02115

Phone: 617-938-7156

Email: [email protected]

*************************************************

[Mike Lacy]

Perhaps you have something fancier in mind, but presuming that by "population" you mean the individuals in your study sample, what you describe sounds like:

logit y x1 x2 .....
predict p
recode p (0/0.04 = 1 "low") (0.04/0.15 = 2 "intermediate") (0.15/1 = 3 "high"), gen(riskcategory)

Whether this categorization is optimal for some clinical or scientific purpose is another question.

[Panagiotis Antiochos]

Dear Mike,

Thank you vey much for your valuable reply.

That perfectly answers my question, i was able to solve it like that.

Kind regards