纵向数据模型均值参数和方差参数的影响分析

Influence Analysis for Mean and Covariance Parameters in Longitudinal Data Model

本文主要研究了纵向数据模型的均值参数和方差参数的统计诊断问题.基于均值参数的加权最小二乘估计和方差参数的REML估计,我们定义了两类参数的广义Cook距离,并求得了计算公式,同时我们研究异常个体的检验问题,得到了两个检验统计量.最后通过实际例子证实了我们的方法的有效性.

In this paper, we consider the diagnostics for the mean and covariance parameters in longitudinal data model. Basing on the WLS (weighted least square) estimators of the mean parameters and REML (restricted maximum likelihood) estimators of the covariance parameters, we defined two types of generalized Cook's distance, and develop the corresponding formulae, moreover we study the problem of outlier subject tests, and provide two test statistics. The analysis methods are illustrated using a real dataset.