Asymptotic properties of penalized splines for functional data (Q2203625)

In the frame of functional data analysis, one concentrates on the asymptotic properties of penalized estimators. There is a theoretic study of the \(L_2\) and uniform convergence of penalized spline estimators for estimating the mean and covariance functions under general functional data settings. Both, fixed common design and random independent design are considered. The study refers to three different types of functional data: sparse, dense and ultra-dense. The obtained convergence rates include the regression spline type asymptotic rates as well as smoothing spline type asymptotic rates. After a short review of the function data model, the second section is devoted to the presentation of penalized spline estimators. Asymptotic properties of the mean function estimator are described in the third section and asymptotic properties of the covariance function estimator are established in the fourth section. All proofs are presented in the Appendix part. Additional proofs for mean and covariance functions estimation are contained in the Supplement to this article, \url{doi:10.3150/20-BEJI209SUPP;.pdf}.