New Statistical Test for Problems with Two and Three Samples, which is More Powerful than the Wilcoxon and Whitney Tests
G. I. Salov
Institute of Computational Mathematics and Mathematical Geophysics Siberian Branch Russian Academy of Sciences
sgi@ooi.sscc.ru
Keywords: two samples, three samples, homogeneity tests, nonparametric tests, noisy image, detection of objects
Pages: 58-70
Abstract
New nonparametric statistics and a criterion (test) based on them are proposed to check the hypothesis of homogeneity of three and two samples, where one sample contains an even number of elements and, hence, can be divided into two samples, against an alternative hypothesis, which implies that random values of one sample are stochastically greater than random values of two other samples. The test is mainly sensitive to shifts of distributions and is more powerful than the Wilcoxon-Mann-Whitney and Whitney tests, at least for problems with samples from exponential and uniform distributions.
|
