|
Abstract:
|
Linear mixed effects (LME) models are increasingly used for analyses of biological and biomedical data. These models assume multivariate normality, which may not be the case, and then robust estimation approaches are preferable to the maximum likelihood ones. Previously proposed robust methods for LME models assume one dimensional response with very limited covariance structures for the multivariate formulation. In many applications, it is desirable to model repeated correlated multiple responses and allow for a general unknown covariance structure. We consider S-estimators for the multivariate nested LME models based on Tukey biweight and investigate their performance in the simulation study. Proposed methodology is applied to analyze repeated measures of three intercorrelated cholesterol components (HDL, LDL, and triglycerides).
|
