20.1 Introduction

Logistic regression analysis is a statistical modeling method for analyzing categorical response data while accommodating adjustments for one or more explanatory variables or 'covariates'. This method is analogous to linear regression analysis for continuous normally distributed responses (Chapter 10), or to ANCOVA (Chapter 11), which is useful for comparing two or more groups while adjusting for various background factors (covariates). Although all of these methods include covariate adjustments, linear regression and ANCOVA analyze means of numeric response measures; logistic regression analyzes proportions based on categorical responses, most commonly binary responses (e.g., success rates, survival rates, or cure rates).

Historically, logistic regression techniques have been widely used for identifying risk factors associated with disease in epidemiological studies. Logistic regression is also popular for analyzing prospective clinical trials and in identifying potentially important covariates in exploratory analyses of clinical research data.