To conduct hypothesis tests of the null hypothesis H0: ∆ = 0 against either one- or
two-sided alternatives, we need some properties of the null distribution of URSS. For
this purpose, we assume that we have perfect judgement rankings for both the X and Y
ranked set samples. Bohn and Wolfe showed that just as for the SRS setting, the RSS
Mann-Whitney statistics URSS (with perfect rankings) is distribution-free under H0 over
the entire class of continuous distributions F. However, there is a major difference in
the null distributions and how critical values are obtained for the two settings. For the
SRS setting, the mk + nq combined sample X and Y observations are not only mutually
16
independent but they are also identically distributed. Thus it suffices to look at each of
the