Logistic Regression Basics

Abstract

What is regression? What’s the difference between linear and logistic regression? When and how should I use them? While these are common questions when students first encounter modeling procedures, there are very few sources which succinctly summarize the process for the SAS® system. After several years of teaching courses in the use of SAS/STAT® for public health data analysis, we developed a primer to quickly impart a working knowledge of logistic regression to our students. While logistic regression analyses may be performed using a variety of SAS® procedures (CATMOD, GENMOD, PROBIT, LOGISTIC and PHREG), this paper focuses on the LOGISTIC procedure as it is particularly well-suited to the needs of our students. Casting regression as a part of a systematic approach to data analysis, we use examples to demonstrate the LOGISTIC procedure’s basic syntax (MODEL, CLASS, OUTPUT statements), model construction and selection options (FORWARD, BACKWARD, STEPWISE, HIERARCHY), and output interpretation. Where relevant the authors have highlighted parallels between LOGISTIC and the, more familiar, REG procedure used for linear regression. The authors hope this paper will serve as a concise reference for those seeking a rapid introduction to logistic regression in SAS®.