EXACT AND RANDOMIZATION DISTRIBUTIONS OF KOLMOGOROV-SMIRNOV TESTS 2 OR 3 SAMPLES

Autor(en): SCHROER, G
TRENKLER, D
Stichwörter: Computer Science; Computer Science, Interdisciplinary Applications; CRITICAL VALUES; EMPIRICAL DISTRIBUTION FUNCTION; K-SAMPLE TEST; KOLMOGOROV-SMIRNOV TESTS; Mathematics; RANDOMIZATION; SEQUENTIAL PROCEDURE; Statistics & Probability
Erscheinungsdatum: 1995
Herausgeber: ELSEVIER SCIENCE BV
Journal: COMPUTATIONAL STATISTICS & DATA ANALYSIS
Volumen: 20
Ausgabe: 2
Startseite: 185
Seitenende: 202
Zusammenfassung: 
The aim of this paper is to compare several test procedures for the two- or three-sample case. These comprise the Birnbaum-Hall test and the k-sample Smirnov test as described in Conover (1980), for instance. To compute the exact distributions of these test statistics, an algorithm developed by Hedges (1957) is extended in two ways, namely, to cover the presence of ties and to generalize it to the three-sample case. This extension can be used to find the distributions of the Birnbaum-Hall and Smirnov test statistic when there are unequal sample sizes, which fills a gap in the literature. Furthermore, we propose a new test where the test statistic measures the area between empirical distribution functions. To make this procedure feasible, a randomized version is studied. The results of a simulation study are reported comparing the performances of several parametric and nonparametric tests. Especially the effects of location or dispersion alternatives are taken into account as well as the presence of ties and outliers.
ISSN: 01679473
DOI: 10.1016/0167-9473(94)00040-P

Show full item record

Page view(s)

6
Last Week
0
Last month
0
checked on Feb 23, 2024

Google ScholarTM

Check

Altmetric