Kernel Two-Sample Testing via Directional Components Analysis
Researchers have developed a new kernel-based statistical test that improves upon existing methods like Maximum Mean Discrepancy (MMD). This novel approach truncates the spectral decomposition of MMD, focusing on robust leading eigen-directions while discarding noisy components. The method demonstrates superior power and robustness, particularly in high-dimensional and unbalanced datasets, while maintaining strict Type I error control. Additionally, it introduces a computationally efficient parametric bootstrap procedure for approximating critical values, offering a faster alternative to permutation-based methods. AI