From: Stan Hilliard
Date: 10 Dec 2002
Time: 13:50:52
I think that the correctness is a matter of judgement. Here are some ideas for thinking it through and making a reasoned choice.
To design your attribute sampling plan, you should compare the following four aspects and choose a plan that provides a reasonable balance between them:
1) AQL and the producer's risk -- because most AQL lots will be accepted.
2) p' = process history because you can expect to reject lots frequently if AQL is less than p'.
3) RQL and the consumer's risk -- (RQL is also called LQ and LTPD) because if a lot is at the RQL level it is almost certain to NOT be accepted.
4) n because the sample size entails the cost of inspection and destructive testing. You might be able to achieve your goal at lower cost by using sequential sampling. You can find more discussion of this by doing a word search in this forum for "sequential sampling". (without quotes)
The key tool that relates these items is the Operating Characteristic Curve: www.samplingplans.com/usingoccurves.htm