A0595
Title: Variable selection for ultra-high-dimensional generalized spatial partial varying coefficient models
Authors: Jingru Mu - Kansas State University (United States) [presenting]
Abstract: The authors propose a generalized partially linear spatially varying coefficient model (GPLSVCM) to accommodate different data types and allow more flexibility simultaneously. The purpose is to study the estimation and structure identification for ultra-high-dimensional generalized partially linear spatial varying coefficient models. A fast and efficient procedure is proposed for identifying model structure via the group adaptive lasso approach and estimating models via spline approaches. The method is shown to be consistent for model structure selection and estimation. The asymptotic normality for the linear components has also been constructed. Simulation studies are conducted to evaluate the performance and use a real spatial dataset to illustrate the application of the proposed method.
