ABSTRACT
This chapter considers a similar situation, in which the two outcome variables have c>2 $ c>2 $ categories. It provides recommendations for practical analysis of paired c×c $ c\times c $ tables. The chapter introduces two examples of paired c×c $ c\times c $ tables, one example with nominal outcome categories and one example with ordinal outcome categories. It explores notation, the sampling distribution, and relevant null and alternative hypotheses for both nominal and ordinal categories. The chapter describes statistical models and methods for tables with nominal categories. It also considers tables with ordinal categories. The chapter provides a test for the stronger hypothesis of symmetry. It illustrates the effect sizes needed to achieve different power levels with the Stuart test for marginal homogeneity and the McNemar-Bowker test for symmetry. The chapter presents some simple methods that take advantage of the inherent information in the ordinal scale of ordinal variables.
