基于似然函数的纵向数据线性混合模型影响分析

LIKELIHOOD-BASED INFLUENCE ANALYSIS IN LINEAR MIXED MODELS FOR LONGITUDINAL DATA

线性混合模型在纵向数据分析中有广泛应用,应用点删除法研究纵向数据线性混合模型影响分析问题.采用的影响度量是基于似然函数构造的广义Cook距离,推导出两种广义Cook距离并用于实际数据分析和统计模拟分析,实际数据分析和统计模拟分析结果都说明该方法是行之有效的.

Linear mixed models （LMM） have been widely used in longitudinal studies. This paper is devoted to discussion of influence analysis in LMM in terms of case-deletion method. The influence measure employed here is a generalized Cook＇s distance which is based on likelihood. Two generalized Cook＇s distances are presented to detect the influential individual. Real data analysis and simulation study show that our method is efficient for influence analysis.