AS 9100C sec 8.2.4 - " Recognized Statistical Principles" meaning

[bigbirdy]

All,

Hi to all Covers! How is everybody doing? I wonder if somebody can explain what is the meaning of ?recognized statistical principles? in AS9100C section 8.2.4. and what is the intent of this clause.

Thanks

Big Birdy

 

[Mikishots]

It means to use a recognized statistical method for sampling, like the old MIL-STD 105. The intent is to ensure that an organization performs sampling using a method that is statistically valid, rather than just something they've made up without justification.

 

[AirdrieQPA]

to Mikishots: Nice twin otter avatar. I worked for the company building the current Series 400 Twin Otter.

I think you got that correct miki - We were struggling with that very question ourselves. Basically you should be using a sampling method with statistical validation behind it. The mil standards are really good for this kind of thing. You need to search around a bit, there are some very good ones available, both mil std and commerical std. Our problem revolved around the small amount of material we bring in, but if you keep looking you can even find standards that cover small lots. That is what we did, we will find out in January 2013 if we as the knight would say in Indian Jones "You chose wisely". One thing about some of the old military standards is that a lot of the mil specs are no longer supported. So chose wisely.

 

[Mikishots]

to Mikishots: Nice twin otter avatar. I worked for the company building the current Series 400 Twin Otter.

I think you got that correct miki - We were struggling with that very question ourselves. Basically you should be using a sampling method with statistical validation behind it. The mil standards are really good for this kind of thing. You need to search around a bit, there are some very good ones available, both mil std and commerical std. Our problem revolved around the small amount of material we bring in, but if you keep looking you can even find standards that cover small lots. That is what we did, we will find out in January 2013 if we as the knight would say in Indian Jones "You chose wisely". One thing about some of the old military standards is that a lot of the mil specs are no longer supported. So chose wisely.

Yep, ANSI/ASQ Z1.4-2008 is the replacement - the MIL standard was tossed a few years back now; that being said, the differences between the two are pretty minor.

 

[Rickser]

Although the old Mil Std and its replacement are OK, statistically in certian situations, they don't fit. We have adapted "Zero Acceptance Number Sampling Plans, 5th Ed. (Nicholas L. Squeglia)" and it works for us with both large and small lot sizes.

 

[dsanabria]

Although the old Mil Std and its replacement are OK, statistically in certian situations, they don't fit. We have adapted "Zero Acceptance Number Sampling Plans, 5th Ed. (Nicholas L. Squeglia)" and it works for us with both large and small lot sizes.

Agree... it is also the one that DCMA (Defense Contract Management Agency) - DOD uses.

In Addition, Large Aerospace have taken the C=O standards, modify it to meet the requirements of their industry and are also an accepted form of valid statistical Principles - See ASQR by UTC Aerospace companies.

 

[Mikishots]

Although the old Mil Std and its replacement are OK, statistically in certian situations, they don't fit. We have adapted "Zero Acceptance Number Sampling Plans, 5th Ed. (Nicholas L. Squeglia)" and it works for us with both large and small lot sizes.

No statement was made to say that it would. It was presented as an example of a statistical method, in response to the OP.

 

[Bev D]

...are also an accepted form of valid statistical Principles...

There is a difference between accepted, popular and a customer requirement and "valid".

For example, the use of c=0 plans is not a statistical validity issue; it is a perception issue. In fact, the perception that a c=0 plan means there are no defects in the lot (which is how this often gets interpreted especially when used for AQL only specified plans) is NOT statistically valid or practically valid. c=0 plans often give much worse protection than a well specified AQL/RQL plan that allows for a c number greater than zero.

This invalid perception stems from our collective basic statistical illiteracy and c=0 plans merely pander to that lack of understanding...but that's just my opinion.