bined sample values from the smallest to the
largest value, irrespective of the population each
value came from. The relevant test statistic then is
the sum of the rank numbers of one sample. A
very small (or very large) rank sum may indicate
that the values from that population tend to be
smaller (or larger) than the values from the other
population.
Over 50 years ago, Wilcoxon [1] introduced
such a nonparametric two sample test for samples
of equal size. Seven years after Wilcoxon’s publication,
White [2] and Van der Reyden [3] extended
the test to two samples of unequal size. A
test similar to Wilcoxon’s was published by Festinger
[4]. Two years after Wilcoxon’s first publication,
however, Mann and Whitney [5] were the
first to explore the case of unequal sample sizes
and to provide tables of critical rank sums for
relatively small sample sizes. It is predominantly
their publication which has been the basis for the
widespread use of the nonparametric two-sample