Covariance estimation for aggregated functional data

This work deals with the problem of estimating covariance functions for functional data, when the available sample consists of sums of curves, rather than individual observations. That is, the population is divided into subpopulations of known sizes and the observations are the sum of all curves for the individuals in the population. We propose (and compare) different methods based on B-Splines, via simulation and application to a real data set, hailing from the electrical energy distribution context.