Variable Selection in Sparse Ultra High-Dimensional Additive Models with Continuous or Discrete Response Variable

• Activity Number: 174
• Type: Contributed
• Date/Time: Monday, July 30, 2012 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistical Computing
• Author(s): Girly Ramirez*+ and Haiyan Wang
• Companies: Kansas State University and Kansas State University
• Keywords: Additive model ; Nonparametric regression ; Sparsity ; Variable selection

Abstract

Variable selection in high dimensional data with possibly discrete response variables requires immediate attention. Motivated by Least Angle Regression, here we propose a multi-step nonparametric model selection algorithm to select variables in sparse ultra-high dimensional additive models. The variables go through a series of nonlinear dependence evaluation following a Most Significant Regression algorithm. The algorithm can be used with continuous or discrete response variable, and when the predictors are linearly or nonlinearly related to the response. Some theoretical properties of the algorithm will be discussed. Simulation results demonstrate that this algorithm works well. Comparisons with other methods such as NIS and INIS (Fan, Feng, and Song, JASA 2011) will be presented.