R&R difference between QS9000 and ASTM

I learned MSA a month ago. I found that the Reproducibility is smaller than the Repeatability in the calculation of Gage R&R. But in all standards of analytical method such as ASTM and ISO standards, Reproducibility is always bigger than Repeatability. How to explain it?

[Don Winton]

Fanjun,

Reproducibility is always bigger than Repeatability.

This should not be the case. Each should be based on the data collected and the respective variations. Perhaps a re-check of the references is in order. In some cases, variation across operators will be greater than variation across trials and vise versa.

Maybe someone else more knowledgeable in the mechanics of the MSA manual methodology could shed some light on this, but I believe the statistics of ANOVA (on which the R&R is based) would preclude this from happening.

Regards,
Don

I have some ASTM Standard Test Methods on hand such as ASTM D5708-95a , ASTM D2007-93 and so on. In D5708-95a,it says the Repeatability of Vanadium is 4.0 mg/kg,the Reproducibility of it is 19 mg/kg,when the concentration is 100 mg/kg. In D2007-93,the result is Repeatability: 1.3%, Reproducibility: 7.8%.
They all show that Reproducibility is bigger than Repeatability,and my other ASTM methods show the same case.
I think you can find them somewhere and will be kindly so as to give me an explanation.

[Don Winton]

Fanjun,

My response was to indicate this is not ALWAYS the case. I could (and will if needed) provide examples where repeatability is larger than reproducibility. The LTV manual has some of this type.

Advice: Do not use examples as the condition. Variability exists just as much across operators as across trials.

Regards,
Don

Could it be that Fanjun is reading a particular test method, that has been analysed before? I could state that on a particular laser micrometer in my lab, on a particular shaft diameter, that the Equipment variation [Repeatability] is 11.1%, the Appraiser variation [Reproduceability]is 4.5% for a combined 12.03% (real example.) This test would of course vary the next time, but in general if I was writing a specification for the test method here I could state the above percentages?

[Don Winton]

Could it be that Fanjun is reading a particular test method, that has been analysed before?

I think that is the case. Your example is is a good explanation of what I was trying to express.

As I stated, I could also provide others, but I believe your points are well taken and should provide enough explanation.

Regards,
Don