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

the parameter estimates can be requested as part of the output. Click on OK in the
main dialog box to run the program. The output generated is from PROC LOGISTIC. It
is omitted here as it is similar to the output previously shown. A check of the Log
window in SAS shows the code that was used to run the analysis.

Conditional Logistic Regression

A conditional logistic regression is demonstrated using PROC LOGISTIC with the MI
dataset. The MI dataset contains information from a study in which each of 39 cases
diagnosed with myocardial infarction is matched with two controls, yielding a total
of 117 subjects.

The model is stated as follows:

logit PðCHD¼ 1 jXÞ¼b 0 þb 1 SMKþb 2 SBPþb 3 ECGþ~

38

i¼ 1

giVi

Vi¼^1 ifith matched triplet

0 otherwise



i¼ 1 ; 2 ;...; 38

The model contains 42 parameters. The data contains 117 observations. The large
number of parameters compared with the number of observations indicates that an
unconditional logistic analysis will yield biased results.

The SAS procedure, PROC PRINT, can be used to view the MI dataset in the output
window. We first run a LIBNAME statement to access the permanent SAS dataset
(mi.sas7bdat) assuming that it is filed on the C drive:

LIBNAME REF‘C:\’;

PROC PRINT DATA¼REF.MI; RUN;

The output for the first nine observations from running the PROC PRINT follows:

Obs MATCH PERSON MI SMK SBP ECG

1 1 1 1 0 160 1

2 1 2 0 0 140 0

3 1 3 0 0 120 0

4 2 4 1 0 160 1

5 2 5 0 0 140 0

6 2 6 0 0 120 0

7 3 7 1 0 160 0

8 3 8 0 0 140 0

9 3 9 0 0 120 0

612 Appendix: Computer Programs for Logistic Regression