Sampling Lasso quantile regression for large-scale data |
| |
Authors: | Qifa Xu Chao Cai Fang Sun Xue Huang |
| |
Institution: | 1. School of Management, Hefei University of Technology, Hefei, Anhui, China;2. School of Statistics, Shandong Institute of Business and Technology, Yantai, Shandong, China;3. Department of Accounting and Information Systems, Queens College, New York, USA;4. Depart of Statistics, Florida State University, Tallahassee, USA |
| |
Abstract: | With the quantile regression methods successfully applied in various applications, we often need to tackle with the big dataset with thousands of variables and millions of observations. In this article, we focus on the variable selection aspect of penalized quantile regression, and propose a new method Sampling Lasso Quantile Regression (SLQR), which allows selecting a small amount but informative data for fitting quantile regression models. Different from the ordinary regularization methods, this SLQR method performs a sampling technique to reduce the number of observations before applying Lasso. Through numerical simulation studies and real application in Greenhouse Gas Observing Network, we illustrate the efficacy of the SLQR method. The numerical results show that the SLQR method is able to achieve a high-precision quantile regression on large-scale data for both prediction and interpretation. |
| |
Keywords: | Quantile regression Large-scale data Sampling Lasso SLQR |
|
|