Malaysian Journal of Mathematical Sciences, April 2019, Vol. 13(S)
Special Issue: 3rd International Conference on Mathematical Sciences and Statistics (ICMSS2018)
The Performance of Robust Heteroscedasticity Consistent Covariance Matrix Estimator
Sani, M., Midi, H., and Arasan, J.
Corresponding Email: habshahmidi@gmail.com
Received date: -
Accepted date: -
Abstract:
The weighted least squares (WLS) method together with heteroscedasticity consistent covariance matrix (HCCM) estimator is often used to
estimate the parameters of a heteroscedastic regression model when the form of errors structure is unknown. However, WLS based on weight determined by hat matrix suffers much set back in the presence of high leverage points (HLPs) in a data set. Moreover, the use of WLS requires an efficient weighting method that will successfully down weight HLPs. In this paper, we proposed new weighting method based on HLPs detection measure in which the good leverage points are allowed to contribute to the estimation of parameters and the bad leverage points are down-weighted as they are responsible for the deviation of the model fit. In the proposed method we employed modified generalized studentized residuals (MGt) with diagnostic robust generalized potentials based on the index set equality (DRGPISE) termed FMGt on the HCCM estimator. The performance of the proposed weighting method is assessed by generated artificial data set.
Keywords: Ordinary least squares, weighted least squares, linear regression, robust HCCM estimator, high leverage points
Indexing
