A0601
Title: Smoothing parameter selection of circular kernel density estimation
Authors: Yasuhito Tsuruta - The University of Nagano (Japan) [presenting]
Abstract: Nonparametric density estimations, such as kernel density estimation, enable flexible estimations. Therefore, nonparametric density estimations of circular data have been studied extensively for decades. Circular kernel density estimators are affected by the choice of the smoothing parameter. Unfortunately, the optimal parameter, which minimizes the mean integrated square error, depends on a derivative of an unknown density. Therefore, many studies have proposed smoothing parameter selectors. A few studies discuss the asymptotic properties of these selectors. The aim is to investigate some properties of the selectors: least-squares cross-validation and direct plug-in rule. The result shows that the convergence rate of the direct plug-in rule is faster than that of least squares cross-validation. The numerical experiment shows the performance of least squares cross-validation and direct plug-in rule under small samples.
