A1286
Title: Towards the most powerful test statistics for randomization tests
Authors: Zijun Gao - University of Southern California (United States) [presenting]
Abstract: Randomization tests are well celebrated for their model-lean validity, but their power is highly dependent on the choice of test statistic. The aim is to first characterize the most powerful test statistic, assuming oracle knowledge of the nuisance functions that characterize the data-generating process. A method is then proposed to estimate these nuisance functions without sample splitting. The resulting randomization test is valid and particularly powerful for small samples and imbalanced designs.
