Sparse nonparametric model for regression with functional covariate. Issue 4 (1st October 2016)

• Record Type: Journal Article
• Authors: Aneiros, G.; Vieu, P.
• Is Part Of: Journal of nonparametric statistics. Volume 28:Issue 4(2016)
• Page Start: 839
• Page End: 859
• Publication Date: 2016-10-01
• Subjects: Functional statistics -- impact points -- nonparametric modelling -- sparse regression -- variable selection; 62G08 -- 62G20; Nonparametric statistics -- Periodicals; 519.5
• DOI: 10.1080/10485252.2016.1234050
• Languages: English
• ISSNs: 1048-5252

Abstract

ABSTRACT: This paper proposes a fully nonparametric model for regression problems involving an infinite-dimensional covariate in which sparsity is modelled in an additive way. The continuous nature of the variable allows to develop new variable selection procedures. Theoretical results show the improvement, in terms of both rate of convergence and number of predictor variables in the model, that one can get from these approaches. An application to some real curves data set is finally presented, which illustrates the double practical interest of the method: good predictive behaviour and interpretability of the outputs as points of the curves being of most impact.