Influence Analysis for Mean and Covariance Parameters in Longitudinal Data Model

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.