What's the difference between 1.32 and 1.33 Cpk?

[Golfman25]

Is there really a material difference between 1.33 and 1.32 Cpk?

 

[Bev D]

0.01

 

[Bev D]

We’ve answered this question many times now. The Cpk number is an estimate. ALL estimates are ‘wrong’. there is a confidence interval that will estimate the uncertainty in the estimate. If you have a small sample size then there is probably no practical difference. If you have a large sample size but you did not get a representative sample size then the estimate is wrong.

If you are concerned about the ‘predicted defect rate’ difference then that is just fake math and has no meaning whatsoever....

 

[Golfman25]

We’ve answered this question many times now. The Cpk number is an estimate. ALL estimates are ‘wrong’. there is a confidence interval that will estimate the uncertainty in the estimate. If you have a small sample size then there is probably no practical difference. If you have a large sample size but you did not get a representative sample size then the estimate is wrong.

If you are concerned about the ‘predicted defect rate’ difference then that is just fake math and has no meaning whatsoever....

I am just concerned about the "bright line" rule -- 1.33 and above = good. Below 1.33 = bad.

 

[Mike S.]

Well, that darn line exists everywhere, right? Any spec where 1.33 is the minimum allowed, is 1.32 really not gonna work? How about 1.329? Maybe, maybe no one would notice, maybe it depends on other tolerance stack-ups whether it would or not.

In the case of a Cpk, I doubt it would matter from a practical standpoint, but practicality may not be the issue.

It always sucks to be near the edge of a spec.

 

[Golfman25]

Well, that darn line exists everywhere, right? Any spec where 1.33 is the minimum allowed, is 1.32 really not gonna work? How about 1.329? Maybe, maybe no one would notice, maybe it depends on other tolerance stack-ups whether it would or not.

In the case of a Cpk, I doubt it would matter from a practical standpoint, but practicality may not be the issue.

It always sucks to be near the edge of a spec.

Only when people don't use their brains.

 

[Matt's Quality Handle]

Only when people don't use their brains.

Why use your brains when ignorance pays so well?

 

[Bev D]

Here is a link to a blog regarding Cpk and early discussions about it that you might interesting. I Ain't Gonna Teach It by Jim Leonard.

 

[Mike S.]

Only when people don't use their brains.

Even when using my brains, I see nothing good coming from operating near the edge of a spec. If you see it otherwise, please enlighten me, maybe I'm not using my brains enough or I don't have enough to begin with.

 

[Bev D]

Mike S - how is a Cpk of 1.32 near a spec?


a Cpk of 1.32 means that the OBSERVED range of data is well within the specification range. only a theoretically miniscule amount of results will be beyond the specificaiton IF it is a perfect Normal distribution which it never is.