ROBUST TECHNIQUE TO SOLVE MULTICOLINEARITY AND OUTLIER

Abstract

       In this paper, we modified the classical ridge regression (RR) to be more resistant for outliers in Y direction. The procedure for modification is by combining the RR with a high breakdown point and high efficiency robust methods generalized M-estimator (GM) based on robust variance covariance matrix such as MRCD.  The largest advantages of the proposed method is that it has less RMSE and higher efficiency than existing methods to overcome the combined problem of multicollinearity and outliers points. A simulation study referred that the suggested method has the sprier performance around of all existing methods.