False discovery rate approach to dynamic change detection
False discovery rate approach to dynamic change detection
复制标题
DOI：
10.1016/j.jmva.2023.105224
复制
发表时间：
2023-08
期刊：
影响因子：
--
通讯作者：
Lilun Du;Mengtao Wen
中科院分区：
文献类型：
--
作者：
Lilun Du;Mengtao Wen
在多数据流监控中，快速、连续地识别行为偏离规范的个体变得尤为重要。在这样的应用中，流的状态可以在空状态和替代状态之间交替，可能多次。为了平衡检测两种类型的变化的能力，即从空到替代和回到空的变化，我们提出了一个新的多重测试过程的基础上的惩罚版本的广义似然比检验统计量的变化检测。在每个时间点的错误发现率（FDR）被证明是控制在一些温和的条件下的数据流的依赖结构。一个数据驱动的方法来选择惩罚参数。通过仿真和算例验证了该方法的优越性。
In multiple data stream surveillance, the rapid and sequential identification of individuals whose behavior deviates from the norm has become particularly important. In such applications, the state of a stream can alternate, possibly multiple times, between a null state and an alternative state. To balance the ability to detect two types of changes, that is, a change from the null to the alternative and back to the null, we propose a new multiple testing procedure based on a penalized version of the generalized likelihood ratio test statistics for change detection. The false discovery rate (FDR) at each time point is shown to be controlled under some mild conditions on the dependence structure of data streams. A data-driven approach is developed for selection of the penalization parameter. Its advantage is demonstrated via simulation and a data example.