基于高维删失数据的分布式惩罚平均经验欧氏似然  

Distributed Penalized Mean Empirical Euclidean Likelihood based on High-dimensional Censored Data

作　　者：朱彦霖 于海生 ZHU Yanlin;YU Haisheng(School of Mathematics and Statistics,Ludong University,Yantai Shandong,264025,China)

机构地区：[1]鲁东大学数学与统计科学学院,山东烟台264025

出　　处：《德州学院学报》2023年第6期15-22,共8页Journal of Dezhou University

基　　金：山东省社会科学规划研究项目(22CSDJ13)。

摘　　要：提出了一种基于高维删失数据的分布式惩罚平均经验欧氏似然方法(DPMEEL)。解决了经验似然在数据量较大时容易出现结果异常的问题,并且通过引入分布式估计的思想,大大提高了计算效率。通过研究表明,在某些条件下,分布式惩罚平均经验欧氏似然具有Oracle特性、渐进正态性,且其似然比的检验统计量服从卡方分布。模拟研究和实例分析说明了分布式惩罚平均经验欧氏似然具有较好的表现。The paper proposes a distributed penalized mean empirical Euclidean likelihood(DPMEEL)method based on high-dimensional censored data.It solves the problem that empirical likelihood is prone to abnormal results when the data is large,and greatly improves the computational efficiency by introducing the idea of distributed estimation.It is shown through the study that under certain conditions,the distributed penalized mean empirical Euclidean likelihood has Oracle property,asymptotic normality,and the test statistic of its likelihood ratio obeys the chi-square distribution.Simulation studies and case analysis illustrate that the distributed penalized mean empirical Euclidean likelihood has a better performance.

关 键 词：经验欧氏似然 高维 惩罚似然 删失数据 分布式估计 

分 类 号：O212.1[理学—概率论与数理统计]