Title: Impact point selection in semiparametric multi-functional regression
Authors: Silvia Novo - University of A Coruna (Spain) [presenting]
German Aneiros - University of A Coruna (Spain)
Philippe Vieu - University Paul Sabatier (France)
Abstract: A new sparse regression model is proposed in the functional data context, which incorporates the influence of two functional random variables in a scalar response: one of them is included linearly, but trough the high-dimensional vector formed by its discretized observations and, the other one, trough a single-index structure. For this semiparametric model, two new algorithms for selecting impact points in the linear part and for estimating the model are presented. Both procedures take advantage of the functional origin of the linear covariates. Some asymptotic results will support theoretically both methods. Finite sample experiments and a real data application will ensure their good practical behaviour.