ABSTRACT
This chapter explores some nonparametric tests. Generally these tests are not as powerful as the equivalent parametric tests, but they can be used when the usual assumptions about the data coming from normally distributed populations do not hold. These nonparametric tests will be ones suitable for analyzing ranked (ordinal) data. The chapter examines the Wilcoxon and Mann-Whitney U two-sample difference tests, as well as the rank sums test confusingly attributed to Wilcoxon, Mann and Whitney. It discusses many-sample difference tests which will consider the Freidman, Kramer and Kruskal-Wallis tests as well as other tests which test for a ranking trend between treatments: the Page test and the Jonckheere test. The chapter also considers Spearman's rank correlation, and explores the relationship to each other and to other nonparametric tests: the binomial test, the McNemar test, the sign test, chi-square, the Cochran Q test, and the contingency coefficient.
