项目名称: 序贯混合似然比检验和快速变点检测方法及其在控制图中的应用研究
项目编号: No.11271135
项目类型: 面上项目
立项/批准年度: 2013
项目学科: 数理科学和化学
项目作者: 濮晓龙
作者单位: 华东师范大学
项目金额: 60万元
中文摘要: 本项目致力于研究序贯检验和变点检测的难点和热点问题,主要包括四个部分内容:(1)序贯混合似然比检验在单边复杂假设检验问题中的研究;(2)快速变点检测方法的若干研究;(3)复杂参数变点检测方法及性质研究;(4)序贯混合似然比方法和快速变点检测方法在控制图中的应用研究。通过本项目,我们将:(1)建立复杂参数区域平均样本量整体评价标准,提出一类全新的序贯混合似然比检验方法,研究新方法的优良性及最优权重函数选取;(2)深入研究单变点快速检测方法高阶性质,建立多变点及多变量数据、在线数据、观测值非独立的快速变点检测方法;(3)对水质监测等领域常出现的一类复杂参数变点检测问题,提出一套全新的快速变点检测方法,研究新方法的评价标准及优良性质;(4)将混合似然比检验方法和快速变点检测方法应用到控制图中,建立多种新型控制图并研究其性质。本项目力图在难点问题上有突破和创新,在热点问题上取得重大进展和研究成果。
中文关键词: 序贯检验;控制图;多元纵向数据;经验似然;缺失数据
英文摘要: In this project, our aim is to solve a challenge problem of the one-sided composite hypothesis in the sequential hypothesis testing and do some new research in the quickset change detection. First, we will propose a class of new sequential mixture likelihood ratio tests in one-sided composite hypothesis. The criterion of the overall performance of the expected sample size will be developed in the composite parameter region, and the theoretical properties and the best choice of the weight functions will be studied. Second, we will do some further research about the quickest change detection problems, by means of developing some asymptotic theories for the one change point quickest detection, and then establishing the quickest change detection methods for four kinds of data, i.e., the multi-change point data, the multivariate data, the online data and the dependent data. Third, some creative research will be developed on the quickest change detection in composite region. We will propose a new method and establish its efficiency and theoretical properties. Finally,applying the methods of sequential mixture likelihood ratio and quickest change detection on control charts, we will develop some new control charts and investigate their properties.
英文关键词: Sequential test;Control chart;Multiple longitudinal data;Empirical likelihood;Missing data