广义线性模型中不完全数据的参数估计  

Parameter Estimation of Incomplete Data in Generalized Linear Models

作　　者：李海霞[1] 张晓冉[1] 徐玉民[1] 

机构地区：[1]燕山大学理学院,河北秦皇岛066004

出　　处：《甘肃联合大学学报（自然科学版）》2008年第6期11-14,共4页Journal of Gansu Lianhe University :Natural Sciences

摘　　要：讨论了广义线性模型类中离散协变量和离散响应变量数据不完全时,参数极大似然估计的EM算法,文中给出了参数估计的具体表达式和估计的渐进协方差.最后,通过随机模拟说明EM算法的优良性.Incomplete data for the class of generalized linear models is examined, in which incompleteness is due to partially missing covariates and response variable on some observations. Under the assumption that the missing data are missing at random, it is shown that the E step of the EM algorithm for any generalized linear model can be expressed as a weighted complete data log-likelihood when the unobserved variables are assumed to form a discrete distribution with finite range. Expressing the E step in this manner allows for a straightforward maximization in M step, thus leading to maximum likelihood estimates （MLE） for the parameters. Asymptotic variances of the MLE are also derived.

关 键 词：EM算法 广义线性模型 极大似然估计 Newton-Raphson算法 

分 类 号：O212.1[理学—概率论与数理统计]