Categorical Data Analysis Using The SAS® System, 2nd Edition

9.1. Introduction

While the typical logistic regression analysis models a dichotomous response as discussed in Chapter 8, “Logistic Regression I: Dichotomous Response,” logistic regression is also applicable to multilevel responses. The response may be ordinal (no pain, slight pain, substantial pain) or nominal (Democrats, Republicans, Independents). For ordinal response outcomes, you can model functions called cumulative logits by performing ordered logistic regression using the proportional odds model (McCullagh 1980). For nominal response outcomes, you form generalized logits and perform a logistic analysis similar to those described in the previous chapter, except that you model multiple logits per subpopulation. The analysis of generalized ...