An exploratory data analysis in scale-space for interval-valued data. Issue 14 (25th October 2016)

• Record Type: Journal Article
• Authors: Park, Cheolwoo; Jeon, Yongho; Kang, Kee-Hoon
• Is Part Of: Journal of applied statistics. Volume 43:Issue 14(2016)
• Page Start: 2643
• Page End: 2660
• Publication Date: 2016-10-25
• Subjects: Kernel smoothing -- nonparametric regression -- SiZer map -- smoothing parameter -- visualization; Statistics -- Periodicals; 519.5
• DOI: 10.1080/02664763.2016.1142947
• Languages: English
• ISSNs: 0266-4763

Abstract

ABSTRACT: We propose an exploratory data analysis approach when data are observed as intervals in a nonparametric regression setting. The interval-valued data contain richer information than single-valued data in the sense that they provide both center and range information of the underlying structure. Conventionally, these two attributes have been studied separately as traditional tools can be readily used for single-valued data analysis. We propose a unified data analysis tool that attempts to capture the relationship between response and covariate by simultaneously accounting for variability present in the data. It utilizes a kernel smoothing approach, which is conducted in scale-space so that it considers a wide range of smoothing parameters rather than selecting an optimal value. It also visually summarizes the significance of trends in the data as a color map across multiple locations and scales. We demonstrate its effectiveness as an exploratory data analysis tool for interval-valued data using simulated and real examples.