Local bandwidth selectors for functional kernel regression

KURUCZOVÁ, Daniela and Jan KOLÁČEK. Local bandwidth selectors for functional kernel regression. 2017.

Basic information

• Original name: Local bandwidth selectors for functional kernel regression
• Authors: KURUCZOVÁ, Daniela and Jan KOLÁČEK.
• Edition: 2017.

Other information

• Original language: English
• Type of outcome: Conference abstract
• Field of Study: 10101 Pure mathematics
• Country of publisher: Spain
• Confidentiality degree: is not subject to a state or trade secret
• WWW: URL
• Organization unit: Faculty of Science
• Keywords in English: Functional data; nonparametric regression; kernel methods; bandwidth selection
• Tags: International impact, Reviewed

Abstract

Nonparametric regression methods based on kernel smoothing rely on selection of suitable bandwidth parameter in order to minimize the mean squared error. We focus on the functional regression, i.e. the case when the predictor is a functional random variable. In finite-dimensional setting, global selection of bandwidth (same bandwidth is used for all data points) is often sufficient. Due to sparsity of infinite dimensional space, local bandwidth selection (different bandwidth is chosen for each data point) seems to be more suitable approach. In our study, we focus on local bandwidth selection methods. Furthermore, we propose local bandwidth selection method based on penalizing functions. Using simulation studies, we compare local bandwidth selection methods with each other and with their global counterparts.

Links

• GA15-06991S, research and development project: Name: Analýza funkcionálních dat a související témata

Investor: Czech Science Foundation