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

The matching factor is the variable MATCH which is coded 1 for a case and 0 for the
two matched controls for each case.

The code to run the conditional logistic regression follows:

PROC LOGISTIC DATA¼MI DESCENDING;

MODEL MI¼SMK SBP ECG;

STRATA MATCH;

RUN;

The distinguishing feature in terms of SAS syntax between requesting an uncondi-
tional and conditional logistic regression is the use of the STRATA statement for the
conditional logistic regression. The STRATA statement in this example declares
MATCH as the stratified (or conditioned) variable.

In earlier versions of SAS, conditional logistic regression could not be run with PROC
LOGISTIC. Instead, PROC PHREG was used and can still be used for conditional
logistic regression. However, PROC PHREG is primarily used to run Cox proportional
hazard models. Conditional logistic regression cannot be run with PROC GENMOD.

The output for the conditional logistic regression using PROC LOGISTIC follows:

The LOGISTIC Procedure

Conditional Analysis

Model Information

Data Set REF.MI

Response Variable MI

Number of Response Levels 2

Number of Strata 39

Model binary logit

Optimization Technique Newton-Raphson ridge

Strata Summary

Response

Pattern

MI Number of

1 0 FrequencyStrata

1 1 2 39 117

Model Fit Statistics

Criterion

Without

Covariates

With

Covariates

AIC 85.692 69.491

SC 85.692 77.777

2 Log L 85.692 63.491

SAS 613