Gage R&R - what is the number of distinct categories?
by Anonymous on 20th June 2005I have just conducted Gage R & R analysis on data gathered from call centre quality monitoring calibration sessions with the aim of checking that there is repeatability and reproducability between call evaluators. Although the results of the data show P values >0.05 and thus I can accept that evaluators are monitoring consistently, the Number of Distinct categories has only come out at 3. This is concerning as I remember from training that the target is 5 or more but cannot remember why.
Can anyone explain the relivence of number of distinct categories and why it needs to be 5 or more? Does the value of 3 impact on the validity of my result?
Comments
Service Providers
Its called 'A graphical Interpretation of Measurement System Variation.' It also describes how to handle single targets and explains the fallacy of the so-called 'guardband.'
Good luck!
Andy
Thanks Andy - do you know what the title of the paper is?
James,
An easy way to think about it is as follows:
Imagine you have a number of calibration references and each is one measurement distribution (+/- 6s) apart. How many calibration references, or measurement distributions, would you need along an 'x' axis to achieve an r-squared = 95%? The answer is approximately 5. If you have a copy of Minitab, you can easily check it out yourself.
You might also want to consider contacting Sophie on this website and ask her for a copy of an article by Andy Urquhart and Peppe Calcara, which is pending publication.
Good luck,
Andy