|
Activity Number:
|
339
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Tuesday, August 8, 2006 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Statistical Computing
|
| Abstract - #306153 |
|
Title:
|
Local Linear Estimation for Single-index Conditional Quantiles
|
|
Author(s):
|
Zhou Wu*+ and Yan Yu and Keming Yu
|
|
Companies:
|
University of Cincinnati and University of Cincinnati and Brunel University
|
|
Address:
|
2921 Scioto Lane 1005, Cincinnati, OH, 45219,
|
|
Keywords:
|
bandwidth selection ; consistency ; pilot estimator ; single-index models
|
|
Abstract:
|
Quantile regression has recently gained much attention for its ability to model conditional distribution of the response variable given the covariates. This paper concerns single-index models for conditional quantiles, which inherit advantages of such models in mean regression context: the unknown link function allowing flexible curvature and the linear index allowing interpretability of parameters. A local linear approach to estimating such models is proposed by minimizing a double weighted check function. With the pilot estimator assumption together with other mild conditions, we study the large sample properties of estimators for both the nonparametric link function and parametric index parameters. Both real data example and simulation results are provided to illustrate the new methodology.
|
