Measurement Error Requirements - 99% or 99.73% spread to represent the 'full' spread?

[pigeon]

Is it an automotive requirement to use the 5.15 or 6 as multiplying factors on the standard deviation to represent the 99% or 99.73% spread on the measurement error?

In the MSA book chapter I section F (page 61 Third Edition : I don't have ordered 4th edition yet) it says that the ISO/IEC guide to the uncertainty in measurement establishes the coverage as sufficient to report uncertainty at 95% of a normal distribution.

Does this mean that an external ISO/IEC17025 laboratory can work with the 95% on the spread of the measurement error? And if we use the same measurement in-house we have to use 99% or 99.73%?

Strange, but maybe I'm missing something because I'm quite a novice in MSA.

Thanks for your ideas

 

[JAltmann]

Re: Measurement Error Requirements - 99% or 99.73% spread to represent the 'full' spr

For measurement uncertianty per ISO-17025 95% confidence interval is used with a k=2.

The 5.15 and 6 sigma you mention is used with GR&R's. I believe with MSA 3rd edition they switched from 5.15 to 6 sigma in the GR&R calculation.

GR&R and the uncertianty are slightly different animals, without going into to great of depth the GR&R incorporates the uncertianty of measure along with the rest of the measurement systems combined errors.

 

[pigeon]

Re: Measurement Error Requirements - 99% or 99.73% spread to represent the 'full' spr

Yes, you are right. They are different, but I'm still wondering why for the one a 95% confidence interval is acceptable and for the MSA a 99% or 99.73% confidence level has to be used instead of 95% interval.
I'm used to work with 95% confidence intervals in statistics. Do we have to use 99% or 99.73% confidence intervals in MSA statistics for automotive or do we have the choice and is it also allowed in automotive to work with the 95% confidence intervals?