الفهرس 
Title : The performance of the robust estimators in regression models
AbstractOnly 14 pages are availabe for public view
 |  | from | 138 |  |  |
| | | from | 138 | | |
Abstract
In heteroscedastic regression models, the use of least squares covariance matrix estimator would not be appropriate in estimation. The problem is further complicated when the data containing outliers or leverage values or both. So. a method which is robust to obtain reliable inference when heteroscedasticity and outliers are present in the data is needed. In this study, new robust heteroscedasticity {u2014} consistent covariance matrix estimators (RHCCMEs) that deliver more reliable inference are proposed. Moreover, a proposed estimator using Blocked Adaptive Computationally Efficient Outlier Nominators for regression data BACON(Y) method and Modified Maximum Likelihood Huber estimator (MM-estimator) instead of Least Trimmed of Squares (LTS) estimator in Two-Step Robust Weighted Least Squares (TSRWLS) estimator. Finally. two robust high leverage detection measures based on robust Mahalanobis distance that based on Minimum Volume Ellipsoid (MVE) estimator and Blocked Adaptive Computationally Efficient Outlier Nominators for multivariate data BACON(X) method are proposed instead of the Hat matrix (H) in Leverage Based Near Neighbour {u2014} Robust Weighted Least Squares (LBNN-RWLS) estimator. The numerical evaluation showed that quasi-t inference based on new proposed estimators offer a substantial improvement over the existing estimators