ENBIS-15 Pre-Conference Workshop on "Measuring Uncertainty in the GUM (VIM) vs ISO 5725 (ISO 3534) Approaches: Which Statistical Viewpoint to Choose?"

6 September 2015; 09:30 – 12:30; Czech Technical University, School of Mechanical Engineering, Prague (Czech Republic)

This half-day ENBIS-15 pre-conference workshop will be given by Rob Willink (New Zealand) and Franco Pavese (Italy), both from IMEKO TC21.

Background Information and Workshop Structure

A measurement is incomplete without a statement of its uncertainty or likely error. But how are systematic errors in a measurement system to be assessed and represented? And what are the other differences between the typical measurement problem and the textbook estimation problem in statistics? This workshop will describe and examine different approaches to the analysis of the 'uncertainty of measurement'.

The workshop is suitable for statisticians, engineers and managers who are responsible for activities involving the analysis of data during any stage of the processing of technical products, as well as academics interested in the topic.

During the workshop participants will learn the different meanings given to "uncertainty" in quantitative measurements, and to systematic and random measurement error, or to Type A and Type B uncertainty, respectively in the GUM and in the ISO (with the necessary reference to VIM and to ISO 3534, respectively, for the terminology). These come from two different approaches in statistics.

The first, today often called ‘traditional’ is the historical “error approach” using a ‘true value’ as a reference. It is still almost universally adopted by ISO standards and generally used in testing and non-quantitative measurements.

The second is an approach basically born in the metrological fields in the ‘80s of the past century, called “uncertainty approach” not using the concept of true value, and adopted in GUM as a Guideline prepared by the BIPM Committee,  with participation of representatives from a number of International Organisms (BIPM, IEC, IFCC, ILAC, ISO, IUPAC, IUPAP, and OIML). Is is used basically in metrology.

Another possible approach is the Bayesian as opposed to the frequentist (or fiducial and other less common, or non-probabilistic approaches. The GUM Committee is presently working on changing the GUM basic principles to a full Bayesian approach. Particular attention will be given to these issues. A neo-classical alternative will be presented, and the different methods will be illustrated by examples.

The workshop will be divided into three parts. Part 1 will cover introduction to the subject matterss - Robin Willink will talk about GUM, and Franco Pavese about ISO 5725 (parts 1 and 2). Part 2 will cover case studies presented by attendees. Part 3 will be dedicated to discussion.

Workshop Topics

The following topics will be covered in the workshop:

  • the existence of 'true values';
  • the association of variances with systematic errors;
  • the representation of fixed quantities by probability distributions;
  • terminology and notation;
  • case studies of practical interest for the GUM and ISO 5725.

Guidance will be provided as to what kind of uncertainty assessment is best in different situations (studies, teaching, certification). Perspective participants are encouraged to submit their cases in advance to one of the chairs of the Program Committee.

Controversial issues will also be discussed. In particular, the universal case of having to perform corrections for systematic errors due to systematic effects will be treated. This apparently simple issue is prone to misunderstanding, confusion and mistakes. That situation is due to many reasons, e.g. by considering the corrections as exact, or by attaching more confidence that would be the case to intra-laboratory data. The importance of inter-laboratory comparison of results will be justified and stressed. Part of the discussion will be both about the advantages of statistical methods as well as their limitations in practice.

Special Requirements for Participants

Participants are expected to bring with them the GUM (and VIM3) and ISO 5725, especially parts 1 and 2 (and the 3534-2:2006).

The GUM and VIM2 are freely available here. ISO 5725 will be projected during the workshop.

The files containing the workshop slides are available for download at the bottom of this page.

Location

The pre-conference workshops will take place at the Czech Technical University, located at the Charles Square, in room 104.