基于M估计的纵向数据线性混合模型中方差的齐性检验  

Testing for Heteroscedasticity in Longitudinal Mixed Effect Linear Models based on M-estimation

作　　者：孙慧慧[1] 林金官[2] 

机构地区：[1]盐城师范学院数学科学学院,江苏盐城224002 [2]东南大学数学系,江苏南京210096

出　　处：《数理统计与管理》2013年第4期646-657,共12页Journal of Applied Statistics and Management

摘　　要：本文对纵向数据的线性混合模型,用Fisher得分法得到了参数的M估计(稳健估计),给出了其渐近性质,研究了M估计下异方差的Score检验问题,并对检验统计量的功效进行了模拟,最后通过葡萄糖数据的实例说明了本文方法的有效性。In this paper, the Fisher scoring method is applied to get M-estimator （robust estimator） in the mixed effects linear model for longitudinal data. And its asymptotic property is given later. The score tests for heteroscedastic errors based on M-estimator are also studied. Then the properties of test statistics are investigated through Monte Carlo simulations. At last, the methods and properties are illustrated by the grape sugar data example.

关 键 词：M估计 纵向数据 异方差 SCORE检验 

分 类 号：O212.2[理学—概率论与数理统计]