Semiparametric Regression Modeling Using Shrinkage Strategies

• Activity Number: 262
• Type: Contributed
• Date/Time: Monday, August 10, 2015 : 2:00 PM to 3:50 PM
• Sponsor: Section on Nonparametric Statistics
• Author(s): Sara DeVries* and Mohamed Amezziane
• Companies: and Central Michigan University

Abstract

In a regression setting, shrinkage estimation arises when we have prior information about the covariates. Based on the sample information only, the unrestricted model is obtained through nonparametric linear smoothing, while, based on the prior information, the restricted model is supposed to be linear. Under this model restriction, we derive several shrinkage regression model estimators. Since the structure of restricted model is fully determined, the newly introduced estimators are semiparametric in nature. We derive distributional properties of the suggested estimators and compare their performance to those of classical estimators. An extensive simulation study is used to confirm our theoretical findings.