Low Sample Size for Gage R&R




I do have a problem to do low sample size GRR due to only 2-3 samples available. What I am thinking is that we can do type 1 GRR, but not sure whether this is correct way. If it's correct, and how can we evaluate the variation between the inspectors? Currently I let 3 or X inspectors measure 1 parts individually 20+ times and calculate the Cg/Cgk individually and also conduct the hypothesis test among different inspectors. Is that works?


Best Regards

Bev D

Heretical Statistician
Super Moderator
why do you only have 2-3 parts available? (the answer could help us provide better advice)

not knowing anything else about the parts or how they are measured:

*I* would have 2-3 operators measure each of the parts that are available a minimum of 30 times. (you are trying to estimate the measurement error - a standard deviation. SDs are difficult to estimate with any kind of precision unless you have a rather large sample size OR you have many small samples.)

*I* would then compare both the SDs and teh part means of each operator to see if any operator to operator differences exist.
Top Bottom