A0407
Title: Divide and conquer approaches for nonparametric regression and variable selection
Authors: Sapuni Chandrasena - University of Toledo (United States) [presenting]
Rong Liu - University of Toledo (United States)
Abstract: The rapid emergence of massive data with increasing size requests new statistical methods, especially in the fields of nonparametric regression, which is flexible but usually computationally intensive. To overcome the limitations of computing and storage, various distributed frameworks for statistical estimation and inference have been proposed. We study the statistical efficiency and asymptotic properties of the spline estimation for generalized additive models using the divide-and-conquer (DAC) approach. We also provide a variable selection method based on the majority voting procedure. The simulation study strongly supports the asymptotic theory and shows that the DAC approach is much more computational expedient without losing much accuracy.
