偏正态自回归模型的异常值得分检验  

Detecting Outliers in Skew-Normal Autoregressive Model by Score Test

作　　者：刘永辉 陈弘扬 王京 LIU Yong-hui;CHEN Hong-yang;WANG Jing(School of Statistics and Information,Shanghai University of International Business and Economics,Shanghai 201620,China)

机构地区：[1]上海对外经贸大学统计与信息学院,上海201620

出　　处：《数理统计与管理》2024年第5期847-855,共9页Journal of Applied Statistics and Management

基　　金：国家社会科学基金重大项目(22&ZD160)。

摘　　要：本文研究偏正态分布下自回归模型的异常值诊断问题。给出了均值漂移和方差加权这两种异常值模型下得分检验统计量的表达式及其渐进分布,数值模拟证实了该检验方法的有效性。实证部分选取了新冠疫情下雪佛龙股票价格的周数据建立偏正态自回归模型,比较分析了同一模型下得分检验法诊断出的异常值和局部影响分析方法诊断出的影响点之间的差异性。In this paper,we investigate how to detect outliers in autoregressive model under skew-normal distribution by score test.We derive mathematical formulas for the score test statistics and its asymptotic distribution under the mean-shift and variance-weighted model,the effectiveness of the test method is verified by numerical simulation.In empirical study,Chevron weekly shares are chosen to construct skew-normal AR model,the differences of effectiveness for detecting influence points and detecting outliers between local influence and score test are compared.

关 键 词：自回归模型 异常值检验 均值漂移模型 方差加权模型 得分检验 局部影响分析 

分 类 号：O212[理学—概率论与数理统计]