Variable Selection in Kernel-Based Nonparametric Regression

• Activity Number: 399
• Type: Invited
• Date/Time: Tuesday, August 6, 2013 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistical Learning and Data Mining
• Author(s): Len Stefanski*+ and Kyle White and Yichao Wu
• Companies: North Carolina State University and NC State University and NC State University
• Keywords: Attenuation ; Kernel regression ; Measurement error ; Model selection ; Shrinkage

Abstract

We start with a review of the measurement-error-model based variable selection framework proposed by Stefanski, Wu and White (2012), measurement error shrinkage and selection operator (MESSO). The approach is then applied to derive variable selection methods for nonparametric regression. The resulting estimator is shown to be related to kernel-based nonparametric regression estimators. The selection feature of the new method is described in detail. Performance of the new method is studied via a combination of asymptotics, Monte Carlo simulation, and applications to real data sets.