GROUP DIAGNOSTIC MEASURES OF DIFFERENT TYPES OF OUTLIERS IN MULTIPLE LINEAR REGRESSION MODEL

The topic of detection outliers is one of the crucial topics that have been of interest to researchers in many scientific fields. The presence of outliers in the dataset may lead to the breakdown of the estimator of the method in use. The statistical literature has shown that several types of outliers occur according to the type and nature of the data. Therefore, the researchers concentrated on identifying the type of outliers of statistical models by using two diagnostic procedures, individual and grouped. Unfortunately, the first procedure neglects the effect of the phenomenon of (masking and swamping). In contrast, the second procedure has not been able to eliminate this phenomenon ideally but rather reduce the rates of its appearance. This paper seeks to suggest improving one of the well-known group diagnostic methods (DRGP) by using an RMVN location and scale matrix instead of MVE to reduce the effect of (swamping). A newly proposed method denoted as DRGP(RMVN) is tested with a simulation study and real data. The results have shown that the performance of our proposed method is more efficient than (DRGP.MVE) to reduce the swamping points.