2019/08/18 by Jun Zhao, Zhao, Jun, Guanao Yan +3
Computer Science · Mathematics · #Bayesian Methods and Mixture Models #FOS: Computer and information sciences #FOS: Mathematics #Methodology (stat.ME) #Statistical Methods and Bayesian Inference #Statistical Methods and Inference #Statistics Theory (math.ST)
paper · pdf · doi:10.48550/arxiv.1908.06431
openalex publication_date 2019/08/18 · openalex created_date 2019/08/22 · openalex updated_date 2026/07/28
Recently, high-dimensional heterogeneous data have attracted a lot of attention and discussion. Under heterogeneity, semiparametric regression is a popular choice to model data in statistics. In this paper, we take advantages of expectile regression in computation and analysis of heterogeneity, and propose the regularized partially linear additive expectile regression with nonconvex penalty, for example, SCAD or MCP for such high-dimensional heterogeneous data. We focus on a more realistic scenario: the regression error is heavy-tailed distributed and only has finite moments, which is violated with the classical sub-gaussian distribution assumption and more common in practise. Under some regular conditions, we show that with probability tending to one, the oracle estimator is one of the local minima of our optimization problem. The theoretical study indicates that the dimension cardinality of linear covariates our procedure can handle with is essentially restricted by the moment condition of the regression error. For computation, since the corresponding optimization problem is nonconvex and nonsmooth, we derive a two-step algorithm to solve this problem. Finally, we demonstrate that the proposed method enjoys good performances in estimation accuracy and model selection through Monto Carlo simulation studies and a real data example. What's more, by taking different expectile weights α, we are able to detect heterogeneity and explore the entire conditional distribution of the response variable, which indicates the usefulness of our proposed method for analyzing high dimensional heterogeneous data.