Logistic Regression: A Self-learning Text, Third Edition (Statistics in the Health Sciences)

whether the subject actually had a knee fracture: Five predictor variables will be used
to obtain predicted probabilities from a logistic regression for each individual in the
dataset. The model follows:

logit PðFRACTURE¼ 1 jXÞ¼b 0 þb 1 AGECATþb 2 HEADþb 3 PATELLAR

þb 4 FLEXþb 5 WEIGHT

The code to run this model is:

Logit fracture agecat head patellar flex weight, or

The output follows:

fracture Odds Ratio Std. Err. z P>jzj [95% Conf. Interval]

agecat 1.743647 .6964471 1.39 0.164 .7970246 3.814567
head 1.243907 .4678455 0.58 0.562 .595172 2.599758
patellar 1.871685 .6584815 1.78 0.075 .9392253 3.729888
flex 1.695114 .6345218 1.41 0.159 .8139051 3.530401
weight 4.50681 1.844564 3.68 0.000 2.020628 10.05199

Directly after running this model an ROC curve can be generated by using thelroc
command. The code and output follows:

lroc

Logistic model for fracture

Number of observations¼ 348

Area under ROC curve¼0.7452

1.00

0.75

0.50

0.25

0.00

0.00 0.25

Area under ROC curve = 0.7452

0.50

1 – specificity

Sensitivity

0.75 1.00

STATA 655