Asymptotic distribution of two-sample empirical \(U\)-quantiles with applications to robust tests for shifts in location (Q764479)

This paper investigates the asymptotic distribution of general two-sample \(U\)-statistics and empirical \(U\)-quantiles based on dependent data that can be represented as functionals of absolutely regular processes. Two scenarios are considered: when there are two independent samples but there is dependence within the samples, and when there is a single process that is a functional of an absolutely regular process and is separated into two disjoint segments. The latter case is the basis for applications to change-point problems. The asymptotic results are applied to construct a robust two sample test for a difference in location between samples of weakly dependent data. The test is based on the median of all pairwise differences. The new test has similar large sample performance to the Wilcoxon-Mann-Whitney test but offers better power when there are heavy tails, skewness or outliers, particularly when one sample is not very large.