Suggested Method for Treatment of Outliers in Panel Data

Abstract

The research deals with the study of the effect of outliers on the results of estimating panel data models, and aims to find ways to diagnose and treat outliers where the box Plot method is used to detect outliers in all variables (independent and dependent) In addition, to confirm the previous diagnostic method, hat Matrix elements have been used to detect outliers in independent variables as well as Mahalanobis distance, while in the dependent variable the outliers have been detected by the Studentized Deleted Residual, After considering several ways to diagnose outliers, it is necessary to measure their effect on the estimated dependent variable values, model coefficients and standard errors as well as measure the effect on all regression coefficients using some scientific methods including (DFFITS,DFBEATS, COOKS DISTANCE, COVRATIO)، To eliminate the effect of outliers on panel data models, where used a Suggested method for treatment outliers, efficient and easily applicable outliers treatment method was used to detect the extent to which outliers affect panel data model tests before and after treatment.