ِالمشاهدات غير المألوفة في نموذج الانحدار

Abstract

The existence of unusual observation ( outliers and leverage points) affects regression models and when these points are real they can not be omitted they must be involved in analysis. Thus, I discus this problem to show how to diagnose this kind of points and to display the influence on regression parameter and the estimated model .This paper studies the properties of the important matrix H (hat matrix) that is useful to diagnose the leverage points .In the application part ,I chose a random sample from a population where such points exist as the researcher think. Then the model is estimated by least square method and the leverage and outlier are diagnosed and their influence on the model is shown .