The Effects of Covariance Structures on the Influence Analysis in Linear Mixed Models

摘要

Influence analysis is an important process of statistical diagnostics. Existing methods of influence analysis in linear mixed models mostly focus on the fixed effects underlying the framework of giving the covariance structures. In this paper, we discuss the effects of covariance structures on the influence analysis. Our study implies that mis-specification of covariance structures may lead mis-specification of influential subjects.