An RKHS Framework for Sparse Functional Varying Coefficient Model

Abstract

We study functional varying coefficient model in which both the response and the predictor are functions of a common variable such as time. We demonstrate the estimation of the slope function for the case of sparse and noise-contaminated longitudinal data. So far, a few methods have been introduced based on varying coefficient model. To estimate the slope function, we consider a regularization method using a reproducing kernel Hilbert space framework. Despite the generality of the regularization method, the procedure is easy to implement. Our numerical results show that the introduced procedure performs well in some senses.