Feature Selection for Ultrahigh Dimensional Varying Coefficient Models

This article is concerned with feature selection for varying coefficient models with ultrahighdimensional covariates. We propose a two-stage approach for these models. The two-stage approach consists of (a) reducing the ultrahigh dimensionality by using a new feature screening procedure based on partial correlation coefficient and (b) applying regularization methods for dimension-reduced varying coefficient models to further select important variables and estimate the coefficient functions. We illustrate the proposed two-stage approach by simulation study and a real data example.

Keywords:

Big data , feature selection , high-dimensional data , rank regression , screening , varying coefficient model