Functional Data Analysis with PACE Package

• Activity Number: 219
• Type: Invited
• Date/Time: Monday, July 30, 2012 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistical Computing
• Author(s): Kehui Chen*+
• Companies: University of California at Davis
• Keywords: functional data analysis ; longitudinal data analysis ; functional regression ; functional principal component analysis ; PACE package

Abstract

This talk will focus on background, methodology and illustrations of the PACE (Principal Component Analysis by Conditional Expectation) package, which is available in Matlab and R and is based on the conditional approach to Functional Data Analysis (FDA) and Empirical Dynamics. It does not rely on pre-smoothing of trajectories, which is problematic when functional data are sparsely sampled or are contaminated with noise. PACE targets estimation and basis representation for both sparsely and densely sampled random trajectories and their derivatives. It implements methodology for functional principal component analysis, curve warping, functional correlation, functional quantiles, functional linear, polynomial, and additive regression and regression diagnostics. The package also includes modules for the analysis of variance processes and for dynamic modeling with empirical differential equations. We will discuss some of the key techniques that underpin PACE and will illustrate use of this package with case studies that involve longitudinal and functional data.