Regression modelling with fixed effects - missing values and other problems

The paper considers new devices to predict the response variable
using a convex target function weighting the response and its
expectation. A MDEP-matrix superiority condition is given
concerning BLUE, RLSE and mixed estimator where the latter is used
in case of imputation for missing values. A small simulation study
compares the alternative estimators. Finally the detection of
non-MCAR processes in linear regression is discussed.