A1395
Title: A knot selection algorithm for regression spline
Authors: Tzee-Ming Huang - National Chengchi Univerisity (Taiwan)
Cheng-Kuan Lin - National Chengchi University (Taiwan) [presenting]
Abstract: In nonparametric regression, spline regression is a commonly used approach for approximating an unknown function. However, to use splines for function approximation, it is necessary to specify knots and the order, among which the selection of knot locations plays a crucial role in determining the quality of the approximation. The proposed algorithm is based on a test statistic designed to detect differences in regression coefficients, and utilizes the outcome of this test to determine the locations of the knots. Both theoretical analysis and numerical experiments demonstrate that this method outperforms the conventional approach based on equally spaced knots.
