Preprint: Fuzzy P-values and Ties in Nonparametric Tests

University of Minnesota, Twin Cities     School of Statistics

Title: Fuzzy P-values and Ties in Nonparametric Tests

Author: Charles J. Geyer

Affiliation: School of Statistics, University of Minnesota

Abstract: This article introduces the use of fuzzy P-values (Geyer and Meeden, 2005) with classical nonparametric tests, specifically with the sign test and the one-sample and two-sample Wilcoxon tests, although the ideas it contains can be applied to any test with a discrete test statistic. It provides simple exact fuzzy tests at any level in the presence of ties.

Key words: sign test; Wilcoxon rank sum test; Wilcoxon signed rank test.

Complete text of the preprint report as PDF. The first version of the paper (which is of no further interest) is still available as PDF.

The R package fuzzyRankTests was written to do the calculations and plots described in the paper. The design document for that package contains detailed proofs omitted from the paper.