Influence analysis on linear models with random effects

In this paper, a unified diagnostic method for linear models with random effects based upon the joint likelihood given by Robinson (in 1991) is presented. The case deletion model is equivalent to mean shift outlier model, as well as case weights model. From this point of view, several new diagnostic measures, such as Cook distance, WK diagnostics are derived. Some previous results are improved. Numerical examples illustrate the method is available. The research supported by NSFC (19631040) and NSFJ.