Prediction Interval for Shrinkage Tuning Parameter

• Activity Number: 196
• Type: Contributed
• Date/Time: Monday, August 10, 2015 : 10:30 AM to 12:20 PM
• Sponsor: International Chinese Statistical Association
• Author(s): Wenhao Hu*
• Companies: North Carolina State University
• Keywords: Lasso ; BIC ; Bootstrap ; Prediction Interval

Abstract

Tuning parameter selection is a challenging problem in the application of selection and shrinkage regression methods. Typically methods based on AIC, BIC, or cross-validation are used to determine a single tuning parameter. Additional variability introduced in to parameter estimates by this selection process is typically ignored. As a step toward characterizing this uncertainty we propose a bootstrap estimator of the sampling distribution of a data-driven tuning parameter. We show that the proposed method provides asymptotically correct coverage and illustrate its finite sample performance in a series of simulation experiments.