A0737
Title: Partially linear models for functional data
Authors: Ming-Yueh Huang - Academia Sinica (Taiwan) [presenting]
Abstract: A class of partially linear models is introduced to predict a scalar response using functional covariates. Different from existing approaches, the models allow the non-linear part to be fully nonparametric. To avoid the curse of dimensionality, we further apply a dimension reduction technique to detect detailed structures of the non-linear part. To estimate the introduced model, an iterative method is proposed. In this method, a novel gradient-based estimation is proposed to estimate the dimension reduction subspace of the non-linear part. We will also discuss the special cases when some of the covariates are multivariate.