Abstract & Overview
Influence diagnostics for detection of influential data points are widely used in statistical modeling to gauge out their impact on various aspect of the analysis. We proposed a method for detection of influential observations in mixed linear model using MINQUE (1) for estimation of variance components and linear unbiased prediction (LUP) for prediction of random effects. The method is based on the analogue of the Cook distance statistics for detection of influential observations affecting both the fixed and random components of a mixed linear model. The method is illustrated with both simulated and the real data sets based on an experimental model for genetic analysis.