Wilcoxon-Mann-Whitney检验：无群体歧视假设
计算机视觉与模式识别
2025-11-26 v1
摘要
传统Wilcoxon-Mann-Whitney（WMW）原假设范围过于宽泛。实际上，WMW检验的是更窄的。推导了标准化统计量（即经验AUC）在正确下的渐近分布，并给出有限样本偏差校正。传统的随机主导性备择假设范围过小。WMW在下一致，Van Dantzig于1951年已证明。
引用
@article{arxiv.2511.20306,
title = {TaCo: Capturing Spatio-Temporal Semantic Consistency in Remote Sensing Change Detection},
author = {Han Guo and Chenyang Liu and Haotian Zhang and Bowen Chen and Zhengxia Zou and Zhenwei Shi},
journal= {arXiv preprint arXiv:2511.20306},
year = {2025}
}