Quantile Regression Approach to the Location-Scale Model

• Activity Number: 36
• Type: Contributed
• Date/Time: Sunday, August 1, 2010 : 2:00 PM to 3:50 PM
• Sponsor: Section on Nonparametric Statistics
• Author(s): Hyokyoung (Grace) Hong*+
• Companies: Baruch College, CUNY
• Keywords: Box-Cox power transformation ; Location-scale model ; Quantile regression ; Longitudinal Study on Aging

Abstract

In this paper, we consider the transformed linear quantile regression methods for location-scale models. When errors are not homoskedastic but variability across different range of quantiles is not so apparent then the location- scale model can offer a good alternative to the quantile regression model since it is more parsimonious and easier to interpret than the quantile regression model. We propose a simple three- step estimation procedure and establish its consistency under some regularity conditions. The simulation studies are shown to examine the performance of the proposed model as well as its competing models. Finally, we present empirical applications in a number of examples.