A1424
Title: Fast non-parametric test on the equivalence of multivariate empirical distributions
Authors: Johannes Bleher - University of Hohenheim (Germany) [presenting]
Luca Sartore - National Institute of Statistical Sciences (United States)
Abstract: A novel approach is presented for testing the equivalence of two multivariate samples using empirical characteristic functions. Building upon an existing framework in the literature, a test statistic is developed that leverages the unique properties of characteristic functions to detect differences in distribution between two multivariate datasets. The method extends an existing approach, which focuses on goodness-of-fit tests for fully specified theoretical distributions. The approach offers a powerful and flexible tool for various applications in statistics and data analysis. The asymptotic distribution of the proposed test statistic is derived under the null hypothesis, and its power is investigated against a range of alternative hypotheses. Through extensive Monte Carlo simulations, the validity of the test performance is investigated. Especially in scenarios involving high-dimensional data or non-normal distributions, the test statistic may offer a computationally efficient way to test for different distributions. Practical guidelines are also provided for implementing the test. Finally, the approach is illustrated through real-world case studies in finance and biostatistics, showcasing its potential for detecting distributional discrepancies in multivariate datasets.
