Algorithm 516: An Algorithm for Obtaining Confidence Intervals and Point Estimates Based on Ranks in the Two-Sample Location Problem [G1]

Abstract

Suppose we are sampling from two populations which have the same continuous distribution except for a possible shift in location. For notational convenience let X1, . . ., X,~ be a random sample from F(x) and let Y1, , Y~ be a random sample from F(x --8). In testing hypotheses concerning 0 the robust two-sample rank test (Wilcoxon-Mann-Whitney) is often used, but the related point estimate and confidence intervals are used relatively rarely. The reasons for this are primarily computational. While it has been practical to carry out the test for large sample sizes, it has not been practical for the estimates.