- Publisher: Al-Qadisiyah Journal for Administrative and Economic Sciences
- DOI: 1816-9171
- Published: May 9, 2021
Robust Technique to Remedy of Unequal Variance Problem in the
Presence of high leverage points
Mohammed A. Mohammed
Department of Materials Management Techniques/
Al-Dewanyia Technical Institute /Al-Furat Alawsat Technical University
|Corresponding Author: Mohammed A. Mohammed Email: firstname.lastname@example.org|
|Abstract : An important assumption of linear regression model is that the variance of disturbances everywhere is
equal (constant variance). However, unequal variance called heteroscedasticity does not cause biasness in estimates,
but it leads to an efficient problem and the standard errors of observations will be inaccurate. Under
heteroscedasticity problem, the ordinary least squares estimates (OLS) are inefficient due to it gives same weights to
all observations regardless of the fact that those with large residuals contain less information about regression
model. The weighted least square (WLS) is a common method for remedy the heteroscedasticity problem.
Unfortunately, in the presence of high leverage points (outlier in the predictor variables), the estimates of classical
method such as OLS and WLS will be damaged and an inefficient. In order to tackle the combined problem of
heteroscedasticity and high leverage points, we suggested a new estimation method called robust quintile weighted
least squares (RQWLS). The results of real data example and simulation study shows that the suggested method has
good performance compared with the existing methods.
Keywords: OLS, WLS, heteroskedasticity, high leverage points, quintile regression, robust standard error.