Variable Selection in Generalized Additive Mixed Models

• Activity Number: 29
• Type: Contributed
• Date/Time: Sunday, July 29, 2012 : 2:00 PM to 3:50 PM
• Sponsor: Section on Nonparametric Statistics
• Author(s): Ahmet Sezer*+ and Cüneyt Toyganözü
• Companies: Anadolu University and Suleyman Demirel University
• Keywords: Variable selection ; semiparametric mixed models ; Penalized Quasilikelihood

Abstract

Identifying the subset of the important variables is of special importance in multivariate regression. In this study we are interested in selecting significant covariates in semiparametric mixed modelling. Variable selection procedure considers both nonparametric and parametric component. We approximate nonparametric component by polynomial splines and minimize the sum of squared errors subject to an additive penalty of spline functions. Semiparametric generalized likelihood ratio test is proposed to select significant variables in the nonparametric component. We show that semiparametric likelihood ratio test statistics follow Chi-squared distribution. Monte Carlo simulation studies are also conducted to examine the finite sample performance of the proposed procedure