首页|期刊导航|山东理工大学学报(自然科学版)|基于e值的r阶向上向下检验过程

基于e值的r阶向上向下检验过程OA

Step-up-down testing procedure of order r based on e-values

中文摘要英文摘要

为解决传统的p值多重检验过程在不规则模型下计算困难问题,在e-BH过程的基础上,提出了基于e值的r阶向上向下检验过程,以进一步提高多重检验过程的检测能力.理论研究该检验过程性能发现,对满足任意相关性条件的e值检验统计量都能将假发现率(False Discovery Rate,FDR)控制到指定水平,同时该过程具有无偏性.使用与FDR、假阴率(False Negatives Rate,FNR)相关的指标,通过模拟实验将基于e值的r阶向上向下检验过程与e-BH过程、传统的p值多重检验过程进行比较,发现该检验过程具有更强的检测能力,完全可以替代传统的p值多重检验过程.

To address the difficulty of getting p-values in some irregular models during the traditional p-value multiple testing procedure,a step-up-down hypothesis procedure of order r based on e-values was proposed,on the basis of the e-BH procedure to further enhance the power of the multiple testing proce-dure.Firstly,the performance of this testing procedure was theoretically studied,and it was found that it can control the false discovery rate(FDR)to the specified level for any e-values test statistic,while maintaining unbiasedness.Then,some simulations were conducted to compare the step-up-down proce-dure of order r based on e-values with the e-BH procedure and traditional p-value multiple testing proce-dures,using indices related to FDR and false negatives rate(FNR).The simulation results illustrate that our proposed testing procedure is more powerful.The results show that this new testing procedure based on e-values is a powerful alternative to the traditional p-value testing procedure.

邹文豪;杨建萍

浙江理工大学 理学院,浙江 杭州 310018浙江理工大学 理学院,浙江 杭州 310018

数理科学

多重假设检验假发现率假阴率e值无偏性

multiple hypothesis testingfalse discovery ratefalse negatives ratee-valuesunbiasedness property

《山东理工大学学报(自然科学版)》 2026 (3)

59-66,71,9

国家自然科学基金面上项目(12071436)

评论