Title: Nonlinear and kernel regression methods for interval-valued data
Authors: Jeong-Taek Jang - Hankuk University of Foreign Studies (Korea, South)
Kee-Hoon Kang - Hankuk University of Foreign Studies (Korea, South) [presenting]
Abstract: Symbolic data are difficult to represent by single value because each observation object has internal structure and variation. Interval data, which is one of these symbolic data, is given as an interval in which all observation objects are not single values. We introduce a regression model using kernel functions and a method of fitting nonlinear regression models to the interval data. We also propose to apply the local linear regression model using the kernel function to the interval data analysis. Various simulations are carried out according to the distribution of the center point and the range by using each method. When various conditions are considered, it is confirmed that the performance of the model based on the proposed local linear regression is quite good. Also, it can be seen that the proposed method shows better performance in the situation where the nonlinear regression function is hard to be estimated well in the real data analysis.