Partially linear additive quantile regression in ultra-high dimension

TL;DR

This paper introduces a flexible semiparametric quantile regression model for high-dimensional data, allowing for nonlinear effects, variable selection, and robustness to heavy tails, with proven theoretical properties and practical validation.

Contribution

It develops a novel partially linear additive quantile regression approach with nonconvex penalties for ultra-high dimensional data, including theoretical oracle properties and efficient estimation techniques.

Findings

Effective variable selection in ultra-high dimensions.

Robustness to heavy-tailed distributions demonstrated.

Successful application to microarray data.

Abstract

We consider a flexible semiparametric quantile regression model for analyzing high dimensional heterogeneous data. This model has several appealing features: (1) By considering different conditional quantiles, we may obtain a more complete picture of the conditional distribution of a response variable given high dimensional covariates. (2) The sparsity level is allowed to be different at different quantile levels. (3) The partially linear additive structure accommodates nonlinearity and circumvents the curse of dimensionality. (4) It is naturally robust to heavy-tailed distributions. In this paper, we approximate the nonlinear components using B-spline basis functions. We first study estimation under this model when the nonzero components are known in advance and the number of covariates in the linear part diverges. We then investigate a nonconvex penalized estimator for simultaneous variable selection and estimation. We derive its oracle property for a general class of nonconvex penalty functions in the presence of ultra-high dimensional covariates under relaxed conditions. To tackle the challenges of nonsmooth loss function, nonconvex penalty function and the presence of nonlinear components, we combine a recently developed convex-differencing method with modern empirical process techniques. Monte Carlo simulations and an application to a microarray study demonstrate the effectiveness of the proposed method. We also discuss how the method for a single quantile of interest can be extended to simultaneous variable selection and estimation at multiple quantiles.