A Computer Program for Fisher's Exact Probability Test

D.L. Tritchler,D.T. Pedrini

DOI: https://doi.org/10.1177/001316447503500328

1975-10-01

Educational and Psychological Measurement

Abstract:The Fisher test is concerned with data of independent groups and a dichotomous criterion in a fourfold (2 × 2) contingency format. Typically, small samples ( N ≤ 30) are considered. Some tables of critical values of Fisher's test for small samples are available to offset the tediousness of computation. Fisher's test requires determination of a combined probability, that is, the observed set added with all the more extreme (directional) sets. For large samples, an approximation method such as chi square is used. Considered in this paper are a mathematical discussion and algorithm, and a computer program for Fisher's test. The computer program for small and large ( N of about 500) samples is available from the authors. Such technology, in most instances, makes approximation methods unnecessary.

psychology, educational, mathematical,mathematics, interdisciplinary applications

What problem does this paper attempt to address?

The problem that this paper attempts to solve is how to determine the significance of the difference between two independent samples or groups in statistical analysis when the scores of these samples or groups fall into two mutually exclusive categories. Specifically, the author focuses on data in the form of a four - fold table (2×2 contingency table). For this type of problem, the Fisher's exact test is usually used to provide an exact probability value. However, for all samples except the smallest ones, it becomes impractical to calculate the Fisher's exact test by hand, so approximate methods such as the chi - square test are usually chosen. This paper proposes a mathematical discussion, an algorithm and a computer program to implement the Fisher's exact test, which is applicable not only to small samples but also to large samples (N is approximately 500), thus making approximate methods unnecessary in most cases. In this way, this research aims to improve the applicability and efficiency of the Fisher's exact test, especially in the case of large sample sizes.