线性纵向数据模型中多个个体的联合影响诊断

Joint influence for multiple subjects in linear longitudinal models

本文利用个体删除方法对具有纯序列相关的线性纵向数据模型,给出了多个个体对参数联合影响的分析式,并将其化简成相对容易计算的形式,同时讨论了enhancing、reducing及swamping效应.进一步,分析了个人所得税申报数据,发现了单个个体删除方法无法识别的影响个体,验证了多个个体删除方法在寻找影响个体时的有效性,扩大了删除方法的应用领域.

Based on subject-deletion diagnostics,joint influence and swamping effects for multiple subjects in linear longitudinal modelwith pure serial correlation are discussed in this paper and the corresponding computational formula for the analysis of joint influence are given. Finally, knew influential subjects are found based on deletion of multiple subjects in income tax returns data, but these subjects are not influential in single subject-deletion diagnostics, this point shows that these methods are effective and their areas are expanded.