Robust two-sample permutation tests (Q1068487)

In this paper the author proposes a new robust two-sample permutation test (RPT) with data-dependent censoring for testing that the joint distribution of two samples is invariant under permutations. The test is shown to be asymptotically optimal against restricted fixed-width neighborhoods, since no other non-parametric test has a p-value with larger minimum slope over the restricted neighborhoods. The RPT also offers the best compromise between robustness and optimality at the idealized alternative, since no other test with the same gross- error-sensitivity has larger slope at the idealized alternative. Some modifications of the test are proposed for testing the nonparametric null hypothesis against neighborhoods of models that have a shared nuisance location-scale parameter. These nuisance-parameter-free versions of the test are justified for large samples from exponential families, and an example for their use is given.