ABSTRACT
This chapter presents an overview of statistical methods commonly used for the analysis of categorical outcomes, focusing primarily on binary and ordinal outcomes. We begin with an overview of contingency table methods that provide simple and intuitive ways of examining the effect of treatment on a categorical outcome. Next, we describe logistic regression models for binary and ordinal outcomes that offer many of the same advantages of contingency table methods while providing an efficient method for incorporating additional covariates (e.g., blocking factors, indicators of trial sites, baseline characteristics of subjects) in the analysis. Finally, we discuss how regression adjustment for baseline values of the outcome variable can generally yield a discernibly more powerful test of treatment effect in randomized clinical trials.
