Asymptotic Behavior of Weights in the Adaptive Regression by Mixing

• Activity Number: 134
• Type: Contributed
• Date/Time: Monday, August 2, 2010 : 8:30 AM to 10:20 AM
• Sponsor: IMS
• Author(s): Chihche Lin*+ and Yuhong Yang
• Companies: Astellas Pharma Global Development, Inc. and University of Minnesota
• Keywords: Model Combining ; Model Averaging ; Model Selection ; Adaptive Regression

Abstract

When several candidate estimators for the mean function are considered in regression, model selection and model combining methods are both choices for estimation. In model selection, efficiency of AIC (Akaike 1973) and consistency of BIC (Schwarz 1978) are frequently discussed and considered prominent in literature. In model combing, Yang's modified ARM with a screening step has been found to achieve an optimality (Lin 2007). Similar to the consistency in model selection, we pursue the issue if ARM and modified ARM give weight that approaches one for the asymptotically best candidate or the minimal-loss candidate when the sample size increases.