TU Darmstadt / ULB / TUbiblio

Residual Analysis to Compare Measurement Series’ Differences with Significance and Size Measures

Hatzfeld, Christian and Bilz, Johannes and Schlemmer, Sascha and Adolf, Jan-Eric and Gu, Yangyang and Elgner, Steffen and Kupnik, Mario (2017):
Residual Analysis to Compare Measurement Series’ Differences with Significance and Size Measures.
In: IEEE SENSORS, Glasgow, [Conference or Workshop Item]

Abstract

Current methods to compare measurement series are based on the comparison of multiple measurement points or fit parameters. These approaches depend on the quality of the fit and are prone to type I error accumulation, when standard statistical tests are used for evaluation. In this work, we present an approach based on a two-sample Kolmogorov-Smirnov test of residuals of data series to a fit in order to assess measurement series differences. This approach does neither rely on physical accurate model functions nor equal sample sizes or variances. It provides comparable measures for data differences as well as a reliability measure of the assessment. It is validated with a Monte Carlo simulation. Applications include the comparison of measurement series with diverse sampling rates or the comparison of measurements without clearly distinguishable properties.

Item Type: Conference or Workshop Item
Erschienen: 2017
Creators: Hatzfeld, Christian and Bilz, Johannes and Schlemmer, Sascha and Adolf, Jan-Eric and Gu, Yangyang and Elgner, Steffen and Kupnik, Mario
Title: Residual Analysis to Compare Measurement Series’ Differences with Significance and Size Measures
Language: English
Abstract:

Current methods to compare measurement series are based on the comparison of multiple measurement points or fit parameters. These approaches depend on the quality of the fit and are prone to type I error accumulation, when standard statistical tests are used for evaluation. In this work, we present an approach based on a two-sample Kolmogorov-Smirnov test of residuals of data series to a fit in order to assess measurement series differences. This approach does neither rely on physical accurate model functions nor equal sample sizes or variances. It provides comparable measures for data differences as well as a reliability measure of the assessment. It is validated with a Monte Carlo simulation. Applications include the comparison of measurement series with diverse sampling rates or the comparison of measurements without clearly distinguishable properties.

Divisions: 18 Department of Electrical Engineering and Information Technology > Institute for Electromechanical Design
18 Department of Electrical Engineering and Information Technology > Institute for Electromechanical Design > Microtechnology and Electromechanical Systems
18 Department of Electrical Engineering and Information Technology > Institute for Electromechanical Design > Measurement and Sensor Technology
18 Department of Electrical Engineering and Information Technology
Event Title: IEEE SENSORS
Event Location: Glasgow
Date Deposited: 07 Nov 2017 08:41
Export:
Suche nach Titel in: TUfind oder in Google
Send an inquiry Send an inquiry

Options (only for editors)

View Item View Item