Comparing regression methods with non-Gaussian stable errors

Document Type : Original Article

Authors

1 Department of Statistics, Faculty of Mathematics and Computer Science, Amirkabir University of Technology (Tehran Polytechnic), Tehran, Iran.

2 Department of Mathematics and Computer Science, Amirkabir University of Technology (Tehran Polytechnic)

Abstract

Nolan and Ojeda-Revah (2013) proposed a regression model with heavy-tailed stable errors. In this paper we extend this method for multivariate heavy-tailed errors. Furthermore, A likelihood ratio test (LRT) for testing significant of regression coefficients is proposed. Also, confidence intervals based on Fisher information for Nolan and Ojeda-Revah (2013) method, called NOR, and LRT are computed and compared with well-known methods. At the end we provide some guidance for various error distributions in heavy-tailed cases.

Keywords

Main Subjects