文章基本信息

标题：Maximum Likelihood Approach for Longitudinal Models with Nonignorable Missing Data Mechanism Using Fractional Imputation
本地全文：下载
作者：Abdallah S. A. Yaseen ; Ahmed M. Gad ; Abeer S. Ahmed 等
期刊名称：American Journal of Applied Mathematics and Statistics
印刷版ISSN：2328-7306
电子版ISSN：2328-7292
出版年度：2016
卷号：4
期号：3
页码：59-66
DOI：10.12691/ajams-4-3-1
出版社：Science and Education Publishing
摘要：In longitudinal studies data are collected for the same set of units for two or more occasions. This is in contrast to cross-sectional studies where a single outcome is measured for each individual. Some intended measurements might not be available for some units resulting in a missing data setting. When the probability of missing depends on the missing values, missing mechanism is termed nonrandom. One common type of the missing patterns is the dropout where the missing values never followed by an observed value. In nonrandom dropout, missing data mechanism must be included in the analysis to get unbiased estimates. The parametric fractional imputation method is proposed to handle the missingness problem in longitudinal studies and to get unbiased estimates in the presence of nonrandom dropout mechanism. Also, in this setting the jackknife replication method is used to find the standard errors for the fractionally imputed estimates. Finally, the proposed method is applied to a real data (mastitis data) in addition to a simulation study.
关键词：longitudinal data; mastitis data; missing data; nonrandom dropout; parametric fractional imputation; repeated measures; standard errors