The Modified Robust Within Group GM-Estimators for the Fixed Effect Panel Data Model

• Main Authors: Nor Mazlina, Abu Bakar@Harun, Habsah, Midi
• Format: Conference or Workshop Item
• Language: English
• Published: 2016
• Subjects: Q Science (General); QA Mathematics
• Online Access: http://eprints.unisza.edu.my/565/1/FH03-FESP-19-22908.pdf; http://eprints.unisza.edu.my/565/
• Institution: Universiti Sultan Zainal Abidin

Block concentrated high leverage points (HLPs) are known to have profound effect on the linear fixed effect regression parameter estimation. They cause heavy contamination and produce bias estimates which lead to wrong analysis and conclusions. Thus, robust regression estimators are introduced to the panel data to provide resistant estimates towards HLPs. Two Robust Within Group GM-estimators (RWGM) are proposed by incorporating two different outlier detection methods; Deleted Robust Generalized Potential (DRGP) and Robust Diagnostic-F (RDF), in the GM-estimator. DRGP and RDF are considered in the study due to their superior abilities to detect outliers correctly in panel data. The performances of the newly proposed methods that we called RWGM-DRGP and RWGM-RDF are studied under two different types of robust centering procedures. The performance of each method is evaluated under Monte Carlo simulations and comparisons are made with the existing RWGM estimator based on Robust Mahalanobis Distances (RMD) by calculating the ratios of root mean square error. The proposed estimators are found to be resilient towards high leverage points due to the success of the weighting schemes by the more superior outlier detection techniques. The results are confirmed through reanalyzing numerical examples.