The two-sample Wilcoxon test generally fails to control the probability of Type 1 error.
We provide a general theory whereby one can construct a permutation test of a parameter class="mathmlsrc">class="formulatext stixSupport mathImg" data-mathURL="/science?_ob=MathURL&_method=retrieve&_eid=1-s2.0-S037837581500138X&_mathId=si2.gif&_user=111111111&_pii=S037837581500138X&_rdoc=1&_issn=03783758&md5=8f7f2e17fef49c4f179df61ccebab7d1" title="Click to view the MathML source">胃(P,Q)=胃0class="mathContainer hidden">class="mathCode">.
The new method controls both the asymptotic probability of the Type 1 error and the exactness property.
The new studentized Wilcoxon test retains all the benefits of the usual Wilcoxon test.
The results are derived for general two-sample class="mathmlsrc">class="formulatext stixSupport mathImg" data-mathURL="/science?_ob=MathURL&_method=retrieve&_eid=1-s2.0-S037837581500138X&_mathId=si1.gif&_user=111111111&_pii=S037837581500138X&_rdoc=1&_issn=03783758&md5=5ec09424ff959a8ed2cc0c3c22437d40" title="Click to view the MathML source">Uclass="mathContainer hidden">class="mathCode">-statistics.