Department of Mathematics
Influence diagnostics and outlier tests for varying coefficient mixed models
In this paper, we consider subset deletion diagnostics for fixed effects (coefficient functions), random effects and one variance component in varying coefficient mixed models (VCMMs). Some simple updated formulas are obtained, and based on which, Cook's distance, joint influence and conditional influence are also investigated. Besides, since mean shift outlier models (MSOMs) are also efficient to detect outliers, we establish an equivalence between deletion models and MSOMs, which is not only suitable for fixed effects but also for random effects, and test statistics for outliers are then constructed. As a byproduct, we obtain the nonparametric "delete = replace" identity. Our influence diagnostics methods are illustrated through a simulated example and a real data set. © 2009 Elsevier Inc. All rights reserved.
"Delete=Replace" identity, Conditional influence, Cook's distance, Influence diagnostics, Joint influence, Outlier tests
Source Publication Title
Journal of Multivariate Analysis
Link to Publisher's Edition
Li, Zaixing, Wangli Xu, and Lixing Zhu. "Influence diagnostics and outlier tests for varying coefficient mixed models." Journal of Multivariate Analysis 100.9 (2009): 2002-2017.