Handling of skewed probability distributions for measurement uncertainties
Reference number | |
Coordinator | RISE Research Institutes of Sweden AB - Mätteknik Borås |
Funding from Vinnova | SEK 860 000 |
Project duration | December 2018 - January 2021 |
Status | Completed |
Purpose and goal
The aim of the project has been to: Develop a mathematical approach for how to handle assymetric uncertainty distributions Improve estimation of measurement uncertainties by taking into account the skewness in uncertainty distributions Increase the knowledge of skewness in measurement uncertainty distributions All these points have been addressed succesfully in the project.
Expected results and effects
The project has resulted in an universal expression for calculation of uncertainty intervals when the relative uncertainty is independent of the measurand level. It includes a parameter that needs to be optimized and that depends on the skewness in the original measurement results. With this expression it is possible to handle a broad spectra of skewness in distributions of mesurement results, where the normal distribution and the log-normal distribution are two ”reference points”. The practical application will be for measurements that have standard uncertainties > approx. 20 %.
Planned approach and implementation
Steps in the project: Development of a transformation facilitating skewed data sets to be transformed to approximate normal distributed data sets Characterization of the transformation Implementation of the transformation on small data sets (in the order of 100 data) Discussion of limitations Implemenation of the transformation The work has mainly been performed by the main applicant and discussed with the participants. A total of 16 group meetings have taken place where results and the report have been discussed. The work has been described in a RISE-report.