Paper detail

Longitudinal Functional Data Analysis

We consider analysis of dependent functional data that are correlated because of a longitudinal-based design: each subject is observed at repeated time visits and for each visit we record a functional variable. We propose a novel parsimonious modeling framework for the repeatedly observed functional variables that allows to extract low dimensional features. The proposed methodology accounts for the longitudinal design, is designed for the study of the dynamic behavior of the underlying process, and is computationally fast. Theoretical properties of this framework are studied and numerical investigation confirms excellent behavior in finite samples. The proposed method is motivated by and applied to a diffusion tensor imaging study of multiple sclerosis. Using Shiny (Chang et al., 2015) we implement interactive plots to help visualize longitudinal functional data as well as the various components and prediction obtained using the proposed method.