Measurement Uncertainty Forum

How do I explain uncertainties that show up on my test data that I supply to the customer.

 
James D Jenkins
Re: How do I explain uncertainties that show up on my test data that I supply to the customer.
by James Jenkins - Saturday, February 11, 2012, 7:32 PM
 
For the record; ALL measurement results are uncertain, in that they are not true values and contain error. This error is unknown in sign or magnitude and can only be dealt with from a containment analysis.

Results should never be reported as "uncertain", as that is a given. Rather, estimating and including information regarding the uncertainty is the thing to do at the minimum. Assuming you are "taking uncertainty into account" in your declaration of "in-tolerance" or "out-of-tolerance", the following is recommended. It is not about certainty or uncertainty, rather it is about confidence with uncertainty taken into account. This can be done many different ways. I would avoid using findings and statements like "indeterminate", again all measurements are indeterminate and uncertain. Rather it is about confidence and probability. Personally, I prefer the focus on "confidence" based on probabilities. So instead of using confusing levels of findings like "indeterminate", "in" or "out", use "in" and "out" of tolerance based on data and add a confidence level finding method.

NOTE to readers: Quametec is working with the organization that submitted this question and have designed a system for them that automatically provides a classification system of confidence levels applicable at the data point level. They are monitoring and communicating with their customers to ensure that the present method provides a value-added understanding of measurement quality and the high level of quality applied by this supplier. We hope to eventually produce a white-paper detailing the techniques and justifications employed.