Nonlinear Regression Modeling and Detecting Change Points via Regularized Basis Expansions

• Activity Number: 637
• Type: Contributed
• Date/Time: Thursday, August 5, 2010 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistical Learning and Data Mining
• Author(s): Shohei Tateishi*+ and Sadanori Konishi
• Companies: Kyushu University and Kyushu University
• Keywords: Change point ; Model selection criterion ; Nonlinear regression ; Regularization

Abstract

We consider the problem of fitting nonlinear regression models to data with change points. In the case that the structure of data is suddenly changed, the use of an ordinary nonlinear regression modeling will lead difficulty of drawing effective information from data. In order to overcome this difficulty, we propose modeling procedures of appropriately estimating nonlinear structure with change points by applying Lasso type of regularization method. We derive a model selection and evaluation criterion from an information-theoretic viewpoint to select tuning parameters in the regularization. We investigate the performance of the proposed nonlinear regression modeling techniques, using real data example and Monte Carlo simulations.