一阶自相关结构下椭球线性混合效应模型的约束推断（英文）  

Restricted Inferences for Elliptical Linear Mixed Models with AR(1) Errors

作　　者：曹春正[1] 任育茜 

机构地区：[1]南京信息工程大学数学与统计学院,江苏南京210044

出　　处：《应用数学》2017年第1期151-161,共11页Mathematica Applicata

基　　金：Supported by the National Natural Science Foundation of China(10671182);the MOE(Ministry of Education in China)Project of Humanities and Social Sciences(13YJC910001)

摘　　要：本文在椭球分布族下研究一阶自相关线性混合效应模型的约束极大似然估计问题.分别考虑位置参数在等式和不等式线性约束这两种情况下的极大似然估计值.同时对约束条件下的兴趣参数给出三种渐近等价的检验方法.蒙特卡洛模拟说明本文方法的有效性和稳健性.本文结合模型对Framingham心脏研究中的胆固醇水平进行了分析.In this paper, we focus on applying restricted maximum likelihood to the linear mixed model with auto-correlated error structure under the elliptical distribution class. Maximum likelihood estimation under both linear equality and inequality constraints on location parameters are considered. Three asymptotically equivalent tests are also constructed for the interest parameters under restrictions. Monte Carlo simulations are conducted to evaluate the effectiveness and robustness of the new methodology. An illustrative example for the cholesterol levels from the Framingham heart study is reported.

关 键 词：约束估计 假设检验 稳健性 纵向数据 自相关 

分 类 号：O212.1[理学—概率论与数理统计]