Title: Distance-based ANOVA for functional data

Authors: Alexandre Homsi Pedott; Flávio Sanson Fogliatto

Addresses: Center of Exact Sciences, Nature and Technology (CENT), UCS, Rua João Del Sasso, 800, Bento Gonçalves, RS, Brazil ' Industrial Engineering Department, UFRGS, Avenida Osvaldo Aranha, 99 Porto Alegre, RS, Brazil

Abstract: We present a method for functional data analysis of variance. Functional data occur when response variables outcomes are a collection of points rather than single values, being usually represented as a profile or curve. The proposed method is an adaptation of the two-way analysis of variance. We used a Euclidean distance as proximity measure between profiles. The method was applied to a simulated gage study based on real industrial data. Results were compared to those obtained using a method available in the literature; in several situations the later led to wrong conclusions about the gage system while the proposed method performed the analysis correctly. [Received 7 May 2014; Revised 25 May 2016; Accepted 27 May 2016]

Keywords: functional ANOVA; functional data analysis; FDA; gauge study; R&R study; repeatability and reproducibility; distance-based ANOVA; analysis of variance; Euclidean distance; simulation.

DOI: 10.1504/EJIE.2016.081018

European Journal of Industrial Engineering, 2016 Vol.10 No.6, pp.760 - 776

Published online: 15 Dec 2016 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article