Object-Oriented Software for Functional Data

TL;DR

This paper presents the funData R package, an object-oriented toolkit for functional data analysis that supports dense, multivariate, and irregular data, facilitating creation, modification, and testing of functional data methods.

Contribution

It introduces a unified, object-oriented framework for functional data analysis in R, including a simulation toolbox and integration with advanced methods like multivariate FPCA.

Findings

Provides a user-friendly, self-contained functional data toolbox

Demonstrates the package's capabilities with real datasets

Enables development and testing of new functional data methods

Abstract

This paper introduces the funData R package as an object-oriented implementation of functional data. It implements a unified framework for dense univariate and multivariate functional data on one- and higher dimensional domains as well as for irregular functional data. The aim of this package is to provide a user-friendly, self-contained core toolbox for functional data, including important functionalities for creating, accessing and modifying functional data objects, that can serve as a basis for other packages. The package further contains a full simulation toolbox, which is a useful feature when implementing and testing new methodological developments. Based on the theory of object-oriented data analysis, it is shown why it is natural to implement functional data in an object-oriented manner. The classes and methods provided by funData are illustrated in many examples using two freely available datasets. The MFPCA package, which implements multivariate functional principal component analysis, is presented as an example for an advanced methodological package that uses the funData package as a basis, including a case study with real data. Both packages are publicly available on GitHub and CRAN.