MSA 4th Edition Variability Study 2 Multiple Readings with P>=2 Instruments- Statistical Help

[HeatQuenchRepeat]

Hello all, I'm new here!

As a result of a recent finding at my company I have been tasked with performing MSA's for a whole slew of evaluation measurement techniques, a few of these tests I have an understanding of and I can build excel templates for, such as the standard Gage R&R, Kappa/Attribute study, and linearity. But I am at a loss on the statistical analysis required for the V2 study per the 4th edition MSA guide chapter IV.

The brick wall that I'm hitting is understanding the Grubbs/Thompsons estimates, my background in statistical analysis is fairly light. Does anyone have any tools/templates that would make my life a bit easier on this or are there any resources that someone can point me to?

I am a fairly new metallurgist, and this is my first dive into MSA's that are not the standard Gage R&R.

 

[Miner]

Have you looked up the reference materials for Grubbs and Thompson? These are on pp 220-221, items 20 and 48. These appear to be behind a paywall.

Are you locked into using the MSA manual for these? If not, check out the Bland-Altman plot or Youden plot. @Bev D has some resources on the latter. I've always used the Bland-Altman plot for comparing two measurement devices. @Bev D uses the Youden plot. They both work well and are easy to understand.

 

[HeatQuenchRepeat]

I found the PDF's for the reference documents on Sci-hub, on cursory examination these seem quite dense but I will dig into them further.

Those analysis techniques are definitely more my speed. I found Bev D's paper and presentation on Youden plots.

Good question, by the letter of the law for IATF 16949 7.1.5.1.1 "Other analytical methods and acceptance criteria may be used if approved by the customer". The question then becomes, if the evaluation measurement technique is for an internal check not for product acceptance for a customer, i.e. dewpoint of an endothermic atmosphere, am I the customer who can then approve an alternate method?

 

[Bev D]

Well there is no such thing as an internal Customer (that’s a myth to discuss later) only paying external Customers. But yes, if good is for internal use only then someone knowledgable should approve it or decide to use it. If you are comfortable with the method and can defend it to others then by all means use it.

As an aside I’ve found that some external Customers don’t understand alternative methods as they are not trained or experienced in practical statistical methods and only know remember exactly what some standard says….

 

[Miner]

They wouldn't understand Grubbs or Thompsons method either.

 

[Bev D]

Youden and Bland-Altman plots drive their eyes into the back of their heads…

 

[HeatQuenchRepeat]

Thanks for the help and feedback y'all!

Seems like I walked into a contentious topic with that internal customer comment