From: Stan Hilliard
Date: 06 Oct 2001
Time: 20:00:06
Hi Bart,
YOU SAID: "Is it current to specify the minimum sample by a "maximum confidence interval? "The controller should take a sample that results in a confidence interval of maximum +- 5%."
MY COMMENT: Here is an example of using confidence intervals for acceptance sampling based on your +- 5%:
EXAMPLE: I determined with TP105 a sample having n=59 and X=0 defectives, the upper one-sided confidence limit for percent defective is 5% (actually 4.9508%). Since the lower limit is zero when X=0, this meets your example of a confidence interval of maximum +-5%.
ANALYSIS: The lower confidence limit for percent good is (100-5)=95. This is below the RQL of 97 percent good specified in your example. So do we reject the lot for not meeting the specification -- even though no defectives occur in the sample? This is not reasonable!!
MY OPINION: Confidence limits are important tools for estimation, but using them for acceptance sampling plans is wrong.
Stan Hilliard