Resampling from centered data in the two-sample problem (Q1116575): Difference between revisions

Bootstrap and permutation approximations to the distribution of U- statistics are shown to be valid when the resampling is from residuals in the two-sample problem. The motivation for using residuals comes from testing for homogeneity of scale in the presence of nuisance location parameters. New asymptotic results for U-statistics with estimated parameters are key tools in the proofs.