D
jager,
Be very cautious of calculations like this.
If customer uses this study to set specifications (and I do hope they verify against VOC), than you may be dismayed to find you cannot achieve the target in production.
Cpk is a statistic and like any statistic it has a sampling distribution.
How many points are in the study?
Is it Cpk (calculated by control chart estimate of sd) or actually Ppk(using rms sd)?
An approximate confidence interval on the Cpk is shown from this link.
http://www.qualitydigest.com/may00/html/lastword.html
More exact intervals can be calculated uding bootstrapping methods.
Note that for samples of 30 points Cpk is +/- 30%. So if you calculate a 1.33 spec limit from a sample of 30, you might be 30% off on a long term basis.
Be very cautious of calculations like this.
If customer uses this study to set specifications (and I do hope they verify against VOC), than you may be dismayed to find you cannot achieve the target in production.
Cpk is a statistic and like any statistic it has a sampling distribution.
How many points are in the study?
Is it Cpk (calculated by control chart estimate of sd) or actually Ppk(using rms sd)?
An approximate confidence interval on the Cpk is shown from this link.
http://www.qualitydigest.com/may00/html/lastword.html
More exact intervals can be calculated uding bootstrapping methods.
Note that for samples of 30 points Cpk is +/- 30%. So if you calculate a 1.33 spec limit from a sample of 30, you might be 30% off on a long term basis.