王健发,肖泽青,王泸怡
(广西大学 数学与信息科学学院,广西 南宁 530004)
广义线性模型中拟似然估计的弱相合性及渐近正态性
王健发,肖泽青,王泸怡
(广西大学 数学与信息科学学院,广西 南宁 530004)
拟似然估计;弱相合性;渐近正态性
广义线性模型(GLMS)由Nelder和Wedderburn[1]引入,作为一种处理非正态响应变量的回归分析模型,其定义如下
其中yi是q维响应变量,Xi是p×q设计矩阵.h∶Rq→Rq是充分光滑的一一映射,β∈Rp是未知的回归参数,β0是它的真值,函数h的逆称为联系函数.给定Xi的条件下,yi( )i=1,…,n独立且有如下的指数分布:
其中:θi称为自然参数;b()·,c()·为相应于指数分布的确定函数.对数似然方程由b(·)和h(·)可求得 ∑(·).在许多情形下假定yi服从指数分布(2)是不切实际的,而且∑(·)的确切表达式不知道.但如果期望(1)假定是正确的,我们仍可用Wedderburm引入的拟似然方法,用拟似然方程
[1]Nelder J A,Wedderburn RWM.Generalized linear model[J].J Roy Statist Soc Ser A,1972,135(3):370-384.
[2]Fahrmeir L,Kanfmann H.Comsistency and asymptotic normality of the maximum likelihood estimator in general⁃ized linear regression[J].Ann Statist,1985,13,342-236.
[3]Chen K,et al.Strong consistency of maximum Quasi-like⁃lihood estimators in generalized linear models with fixed and adaptive design[J].Ann Statist,1999,27,1155-1166.
[4]Yue L,Chen X R.Rate of strong consistency of qua⁃si-maximum likelihood estimate in generalized linear mod⁃els[J].Science in China Series A-Mathematics,2004,47(6):882-893.
[5]赵林城,尹长明.广义线性模型中极大拟似然估计的强相合性[J].中国科学A辑:数学,2005,5(3):312-317.
[6]尹长明,赵林城.广义线性模型中极大拟似然估计的渐进正态性与强相合性[J].中国科学A辑:数学,2005,35(11):1236-1250.
[7]高启兵,吴耀华.广义线性回归拟似然估计的强相合性[J].数学年刊:A辑,2004,25(6):705-710.
[8]Liang K Y,Zeger S L.Longitudinal data analysis using gen⁃eralized linear models[J].Biometrika,1986,73(1):13-22.
[9]张三国,廖源.关于广义线性模型拟似然估计弱相合性的几个问题[J].中国科学A辑:数学,2007,37(11):1368-1376.
[10]戴牧民,陈武华,张更荣.实分析与泛函分析[M].北京:科学出版社,2007:28.
[11]Liao Y.Weak consistency of quasi-maximum likelihood estimates in multivaiate generalized linear models[J].Chi⁃nese Journal of Applied Probability and Statistics,2006,22(3):288-294.
Weak Consistency and Asymptotic Normality of Quasi-Maximum Likelihood Estimates in Generalized Linear Models
WANG Jianfa,XIAO Zeqing,WANG Luyi
(School of Mathematics and Information Science,Guangxi University,Nanning530004,China)
In a generalized linear model withq×1 responses,bounded and fixedp×qregressorsXiand natural link func⁃tion,we study the solutionof quasi-maximum likelihood equationand other conditions,and use contractive mappings to prove that the quasi-likelihood equation has a solutionof weak consistency;Fort>0,let ψ(t) be a positive nondecreasing function such thatis a convex function,which prove the asymptotic normality of maximum quasi-likehood estimatorβn.
quasi-maximum likelihood estimation;weak consistency;asymptotic normality
O 212.1
A
1674-4942(2011)02-0137-04
2011-01-25
国家自然科学基金(11061002)
毕和平