B1081
Title: Kernel-based nonparametric regression for cylindrical data
Authors: Yasuhito Tsuruta - The University of Nagano (Japan) [presenting]
Abstract: Many studies have discussed regression where a predictor has support on a circle and responder has support on real line, such as wind direction and speed. Such data is also called cylindrical data. Kernel-based nonparametric regressions are flexible in estimating the shape of an underlying regression model for cylindrical data. The smoothing parameter plays an important role in determining the shape of the nonparametric regression. Therefore, for the nonparametric model under the specific kernel class, this study derives the optimal smoothing parameter minimizing weighted conditional mean integrated squared errors and the convergence rate. The numerical experiment is conducted to investigate the performances in small samples for the nonparametric regression.