Q. 143  Repeatability is considered a pre-requisite for Accuracy assessment when we deal with Attribute Agreement Analysis but it is not so when we work with Gage R&R for variable (continuous) data. Why is it so? 

 

  

Note for website visitors - Two questions are asked every week on this platform. One on Tuesday and the other on Friday.

• All questions so far can be seen here - https://www.benchmarksixsigma.com/forum/lean-six-sigma-business-excellence-questions/
• Please visit the forum home page at https://www.benchmarksixsigma.com/forum/ to respond to the latest question open till the next Tuesday/ Friday evening as per Indian Standard Time.  
• The best answer is always shown at the top among responses and the author finds honorable mention in our Business Excellence dictionary at https://www.benchmarksixsigma.com/forum/business-excellence-dictionary-glossary/ along with the related term.

 

As we do attribute agreement analysis for attributes data, it measure only repeatability.Which can’t give any precise value. But in continuous data R&R, we can measure precise value with continuous value. So that we can  accept with some close value. It is called accuracy 

Attribute Agreement Analysis typically involves a binary decision making around one or more attributes of the item being inspected i.e. accept or reject, good or bad, pass or fail etc.,  So, in this context we are referring to discrete data type. The key call out here is the human factor (known as appraisers) who typically make these assessments. With this condition in place, it is  all important that the appraisers are consistent with themselves (repeatability), with one another appraisers (reproducibility) and with known standards (team accuracy). 

 

A poor repeatability implies the assessor himself is not clear on the measurement system analysis and hence needs a  thorough training, understanding and alignment else the entire objective of the assessment will not make sense.

 

On the other hand, in Gage R&R, we deal with continuous  data type  and there is a Gage or measuring instrument which plays a significant role to measure the data.

Each appraiser will measure each item multiple times (repeatability), and their average measurement of each item will be compared to the average for the other appraisers or measurement tools (reproducibility). It is primarily the measurement tool that plays primary significance in Gage RandR and what is important is to ensure the equipment calibration is taken care off so as the repeatability factor of appraiser measuring the item is more Gage dependent rather than the appraiser himself.

In Attribute Agreement Analysis, bias needs to be eliminated as a non meaurable parameter is either being accepted or rejected and hence repeatability becomes a prerequisite. 

In the case of Gage R&R, the comparison is a measurable parameter and hence a certain level of variation even if it exists can be allowed.

Two reasons

1. The name itself. In Attribute agreement analysis, all 4 things are calculated (repeatability, appraiser accuracy, reproducibility and team accuracy). On the other hand, Gage R&R is only used for repeatability and reproducibility. There is a different test to check for accuracy.

2. In attribute data, there is only one right answer (e.g good / bad, defective or non defective). Hence if given the same defective product, for an appraiser too be accurate he has to be repeatable. But in variable data, there could be a range which is correct. Appraiser might not be repeat his observations but still be correct as long as he is in the range / tolerance. E.g. the landing gear on the plane can only be lowered below a certain speed (say 200 knots). Now pilot may lower it at 200 or 150 as well. He will not be repeatable but will be accurate

Congratulations to all 4 respondents. You have hit the nail right on the head - one cannot expect absolute repeatability in continuous data because if which it is not a pre- requisite in Gage R&R. However, attribute data has responses which can be counted and categorized leading to absolute repeatability - there are no possible acceptable ranges unlike continuous data.

 

The chosen best answer is that of Natwar Lal as he explains his point with an example. However, to point 1 in the answer - the name of the test is derived from what you are testing and not the other way around.