Linear Quantile Regression Based on EM Algorithm

Abstract

This article aims to put forward a new method to solve the linear quantile regression problems based on EM algorithm using a location-scale mixture of the asymmetric Laplace error distribution. A closed form of the estimator of the unknown parameter vector β based on EM algorithm, is obtained. In addition, some simulations are conducted to illustrate the performance of the proposed method. Simulation results demonstrate that the proposed algorithm performs well. Finally, the classical Engel data is fitted and the Bootstrap confidence intervals for estimators are provided.

Keywords:

• Quantile regression
• EM algorithm
• Asymmetric Laplace distribution (ALD)
• Bootstrap confidence interval

Mathematics Subject Classification:

• 62J05