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

ROC Curve Using Knee Fracture Data
Sensitivity
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
0.0 0.1 0.2 0.3 0.4 0.5
1 – specificity

0.6 0.7 0.8 0.9 1.0

Polytomous Logistic Regression

A polytomous logistic regression is now demonstrated with the cancer dataset using
PROC LOGISTIC. If the permanent SAS datasetcancer.sas7bdatis on the C drive,
we can access it by running a LIBNAME statement. If the same LIBNAME statement
has already been run earlier in the SAS session, it is unnecessary to rerun it.

LIBNAME REF‘C:\’;

First a PROC PRINT is run on the cancer dataset.

PROC PRINT DATA¼REF.CANCER; RUN;

The output for the first eight observations from running the proc print follows:

Obs ID GRADE RACE ESTROGEN SUBTYPE AGE SMOKING

1 10009 1 0 0 1 0 1

2 10025 0 0 1 2 0 0

3 10038 1 0 0 1 1 0

4 10042 0 0 0 0 1 0

5 10049 0 0 1 0 0 0

6 10113 0 0 1 0 1 0

7 10131 0 0 1 2 1 0

8 10160 1 0 0 0 0 0

PROC LOGISTIC can be used to run a polytomous logistic regression (PROC CAT-
MOD can also be used).

The three-category outcome variable is SUBTYPE, coded as 0 for Adenosquamous, 1
for Adenocarcinoma, and 2 for Other. The model is stated as follows:

ln

PðSUBTYPE¼gjXÞ

PðSUBTYPE¼ 0 jXÞ



¼agþbg 1 AGEþbg 2 ESTROGENþbg 3 SMOKING

whereg¼ 1 ; 2

SAS 617