Analysis of structural equation models with censored or truncated data via EM algorithm

被引:8
|
作者
Tang, ML [1 ]
Lee, SY [1 ]
机构
[1] Chinese Univ Hong Kong, Dept Stat, Shatin, NT, Hong Kong
关键词
structural equation models; non-ignorable missing mechanism; generalized censoring mechanism (GCM); maximum-likelihood estimation; generalized EM algorithm; censored data; truncated data;
D O I
10.1016/S0167-9473(97)80040-0
中图分类号
TP39 [计算机的应用];
学科分类号
081203 ; 0835 ;
摘要
In this article, we delineate the analysis of structural equation models with censored data or truncated data with the number of unmeasured subjects under the framework of the generalized censoring mechanism (GCM), one of the most popular non-ignorable missing mechanisms. An EM algorithm for maximum-likelihood estimation in structural equation models with censored data will be presented. By reformulating the censored-data problem into non-ignorable missing-data problem, the maximization procedure is shown to be simplified. A simulated data set will be used as an illustrative example. (C) 1998 Elsevier Science B.V. All rights reserved.
引用
收藏
页码:33 / 46
页数:14
相关论文
共 50 条