Sample size in the field of valve calibration
How to proceed to get the right sample size in order to test for normality and chech sytem capability?
Our process is concerned by calibrating different type of valves installed in many sites all over the country. those valves have different pressure references and could be calibrated using either water or gaz.
in our accreditation project, we need to be sure of the normality in order to use the statistics related and carry on our 17025 study.
Assign topic to the user
I assume the sample size you refer to is for your data points, not related to sampling size of the test items. Typically a rule of thumb has always been 30 data points as a minimum.
You mentioned statistics related to your ISO 17025 study. It really depends what type of study. The number itself is not the only consideration. As the normality assumption is about the distribution of the sample mean, you may, for example need to do a test for outliers before the normality check. There again normal distribution is not always a requirement for study. For purposes of a control chart a sample size of 30 is suitable however for other studies fewer may be suitable whilst others may require more data points.
One point to bear in mind is that statistical significance and significance in providing valid result are not the same thing. The consideration of measurement uncertainty is important for calibrations.
Have a look at
- The response to a question on the concept of measurement uncertainty at https://community.advisera.com/topic/iso-17025-queries/
- Evaluation of Measurement Uncertainty Procedure at https://advisera.com/17025academy/documentation/evaluation-of-measurement-uncertainty-procedure/ This covers the basic principles and steps to plan, measure and calculate the data required for an evaluation of measurement uncertainty. The two appendices related to the document, Measurement Uncertainty Checklist and Measurement Uncertainty Record support the process. I recommend you also look to your sector and suppliers for commonly used approaches.
Comment as guest or Sign in
Dec 12, 2020