Recovering Covariance from Functional Fragments

TL;DR

This paper develops a nonparametric method for estimating covariance functions from fragmented functional data, addressing the challenge of limited observation windows and translating the problem into low-rank matrix completion.

Contribution

It introduces a novel nonparametric estimator for covariance from fragmentary data under smoothness and rank conditions, with precise grid and observation requirements.

Findings

Estimator performs well on simulated data.

Method is effective on real-world datasets.

Conditions for identifiability are explicitly characterized.

Abstract

We consider nonparametric estimation of a covariance function on the unit square, given a sample of discretely observed fragments of functional data. When each sample path is only observed on a subinterval of length $\delta<1$, one has no statistical information on the unknown covariance outside a $\delta$-band around the diagonal. The problem seems unidentifiable without parametric assumptions, but we show that nonparametric estimation is feasible under suitable smoothness and rank conditions on the unknown covariance. This remains true even when observation is discrete, and we give precise deterministic conditions on how fine the observation grid needs to be relative to the rank and fragment length for identifiability to hold true. We show that our conditions translate the estimation problem to a low-rank matrix completion problem, construct a nonparametric estimator in this vein, and study its asymptotic properties. We illustrate the numerical performance of our method on real and simulated data.