Title: Randomization empirical processes and deduced hypothesis tests
Authors: Dennis Dobler - Vrije Universiteit Amsterdam (Netherlands) [presenting]
Abstract: Paralleling results for the bootstrap and random permutation, a broadly applicable conditional Donsker theorem is developed for empirical processes which are based on randomized observations. Random elements of an algebraic group are applied to the raw data and a statistic is applied to the randomized data. After a suitable studentization of the statistic, asymptotically exact hypothesis tests are deduced. They are even finitely exact under group-invariant hypotheses. The methodology is exemplified with a right-censored paired data problem and the question: which entry has the better survival chances? To this end, a suitable Wilcoxon-type test will be developed.