arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2409.15628math.STstat.TH
收藏

基于图的总体变差积分概率度量(TV IPM)的双样本检验

Two-Sample Testing with a Graph-Based Total Variation Integral Probability Metric

Alden Green, Sivaraman Balakrishnan, Ryan J. Tibshirani

AI总结:

针对多元非参数双样本检验问题,提出基于图近似总体变差积分概率度量(TV IPM)的图TV检验方法,证明了其在检测TV IPM中分离的备择假设时具有极小极大风险最优性,并揭示了$\chi^2$检验的次优性。

摘要:

我们考虑了一种新颖的多元非参数双样本检验问题,在备择假设下,分布P和Q在由有界总体变差函数构成的积分概率度量(TV IPM)中被分离。我们提出了一种新的检验方法,即图TV检验(graph TV test),它使用基于图的TV IPM近似作为检验统计量。我们证明了使用$\varepsilon$-邻域图计算并通过置换校准的该检验,对于检测在TV IPM中分离的备择假设具有极小极大风险最优性。作为一个重要的特例,我们证明图TV检验对于检测空间局部化的备择假设是最优的,而$\chi^2$检验被证明是次优的。我们的理论得到了模拟数据和真实数据的数值实验支持。

英文摘要:

We consider a novel multivariate nonparametric two-sample testing problem where, under the alternative, distributions $P$ and $Q$ are separated in an integral probability metric over functions of bounded total variation (TV IPM). We propose a new test, the graph TV test, which uses a graph-based approximation to the TV IPM as its test statistic. We show that this test, computed with an $\varepsilon$-neighborhood graph and calibrated by permutation, is minimax rate-optimal for detecting alternatives separated in the TV IPM. As an important special case, we show that this implies the graph TV test is optimal for detecting spatially localized alternatives, whereas the $χ^2$ test is provably suboptimal. Our theory is supported with numerical experiments on simulated and real data.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑