|
Activity Number:
|
52
- New Challenges in Complex Data Analysis
|
|
Type:
|
Invited
|
|
Date/Time:
|
Sunday, July 30, 2017 : 4:00 PM to 5:50 PM
|
|
Sponsor:
|
Korean International Statistical Society
|
|
Abstract #322008
|
View Presentation
|
|
Title:
|
Wild Residual Bootstrap Inference for Penalized Quantile Regression with Heteroscedastic Errors
|
|
Author(s):
|
Lan Wang* and Ingrid Van Keilegom and Adam Maidman
|
|
Companies:
|
University of Minnesota and ORSTAT, KU LEUVEN and University of Minnesota
|
|
Keywords:
|
quantile regression ;
bootstrap
|
|
Abstract:
|
Penalized quantile regression model provides a useful tool for analyzing heterogeneous data. By flexibly accommodating varying covariate effects at different quantile levels, we may obtain a more complete picture of the conditional distribution of a response variable. By allowing the sparsity level to be different at different quantile levels, we permit a more realistic assumption of sparsity. Although considerable attention has been devoted to penalized quantile regression in recently years, most of the existing work has been focused on point estimation. It is still largely unknown how to attach an accurate standard error to the penalized quantile regression estimator. Bootstrap procedures have been recently studied for penalized mean regression but are not directly applicable to penalized quantile regression with heteroscedastic errors. We consider a wild bootstrap procedure for penalized quantile regression and prove its consistency. Numerical results demonstrates its satisfactory performance. (Joint work with Ingrid van Keilegom and Adam Maidman).
|
Authors who are presenting talks have a * after their name.
