How do I generate a risk score and then use it to calculate quartile odds ratio

[Lena Mathews]

Hello

I was looking for some assistance in generating a prediction risk score for a binary outcome (cr_code), then using the risk score to calculate the quartile odds ratio for patients in the highest risk quartile compared to the lowest risk quartile for having the event.

Variables I would like to include in the risk score are ones that were significant in a logistic regression backwards selection analysis: age, race, gender, income, education cognitive function. This is what I have so far you:

logit cr_code i.income i.education i.q_cognitive2
predict yhat, xb

Your assistance will be greatly appreciated.

[Rich Goldstein]

first, when you use the -predict- command, use the "pr" option not the "xb" option; see

help logit postestimation##predict

there are a number of ways you can divide the resulting risk score into quartiles if that's what you really want; see, e.g.,

help pctile