I made a GR&R to a valve tester, the valve tester has the following test settings:
0 to 850 psi fails 850 to 999 psi is a #1 valve 1000 to 1099 psi is a #2 valve 1100 to 1199 psi is a #3 valve 1200 to 1299 psi is a #4 valve 1300 to 1399 psi is a #5 valve 1400 to 1549 psi is a #6 valve Above 1550 psi fails.
I took 1 or 2 valves from each of the types of valve from #1 to #6 and the results are attached since I couldn't upload pictures to a image hosting website from my work's browser..
The issue here is that if I grab a valve and it measures 1400, then I test it again and it measures 1375, the operator is not going to measure it twice so he is going to label it as wathever the first reading belongs to, I hoped that by performing the GR&R it would show me a high variation % from the equipment but it didn't..
My question is should I have picked 10 samples but only from 1 type of valve rather than 10 samples from all the types of valves ? for instance 10 valves varying from 1300 to 1399 rather than 10 valves from 0 to 1550 ??
There are a number of ways that this could be handled.
First, I will assume that this is strictly an inspection (and classification) device and is NOT used for process control such as SPC. Given that assumption, the method of collecting your samples does not affect the calculation of the %Tol metric, so your approach is acceptable.
While I prefer using the actual numerical results, which you did use, you could also perform an attribute gage study to assess how well the device classifies the valves. However, I believe that the approach taken was the better approach.