Statistical Hypothesis Testing with SAS and R

Part VI

Nonparametric Tests

In this part we cover classical nonparametric tests such as the sign test, the signed-rank sum test and the Wilcoxon test. Further nonparametric tests are found throughout the book and listed with their parametric alternatives such as the Spearman rank correlation coefficient and the Fisher exact test. This arrangement is arbitrary but we think that it serves the intention of this book best. Also, nonparametric tests is a broad term and we refrain from tediously classifying each test according to such criteria.

In SAS you can perform most of the tests presented here with the procedure PROC NPAR1WAY. The procedure also allows to calculate the exact distribution and therefore exact p-values. However, this is very cumbersome and time consuming (even taking days to calculate). Furthermore exact p-values via Monte Carlo estimation can be calculated. We do not present this here and refer the reader to the SAS documentation. In most cases the asymptotic p-values should suffice. In R the tests are covered by single functions, described here.