PubAg

Main content area

Simultaneous estimation of quantile regression functions using B-splines and total variation penalty

Author:
Jhong, Jae-Hwan, Koo, Ja-Yong
Source:
Computational statistics & data analysis 2019 v.133 pp. 228-244
ISSN:
0167-9473
Subject:
algorithms, data collection, regression analysis
Abstract:
We consider the problem of simultaneously estimating a finite number of quantile functions with B-splines and the total variation penalty. For the implementation of simultaneous quantile function estimators, we develop a new coordinate descent algorithm taking into account a special structure of the total variation penalty determined by B-spline coefficients. The entire paths of solution paths for several quantile function estimators and tuning parameters can be efficiently computed using the coordinate descent algorithm. We also consider non-crossing quantile function estimators having additional constraints at the knots of spline functions. Numerical studies using both simulated and real data sets are provided to illustrate the performance of the proposed method. For a theoretical result, we prove that the proposed the quantile regression function estimators achieve the minimax rate under regularity conditions.
Agid:
6195561