From: Stan Hilliard
Date: 07 Jan 2004
Time: 10:14:03
Hi bayanbaru,
I am assuming that you refer to attribute sampling plans.
To specify such a sampling plan, you need to pick AQL, Alpha, RQL, Beta.
(LTPD and LQ and RQL mean exactly the same thing.)
AQL and Alpha together specify the fraction defective (AQL) of a lot that the plan will have a small probability (Alpha) of rejecting. (1-Alpha) = the probability of accepting a lot if it contains AQL fraction defective.
AQL and Alpha define the “producer’s point” of the operating characteristic curve of the plan. (oc curve). It is called the “producer’s point” because it satisfies the producer’s intentions of usually accepting lots if those lots are truly AQL fraction defective.
See www.samplingplans.com/usingoccurves.htm
RQL and Beta together specify the fraction defective (RQL) of a lot that the plan that will have a small probability (Beta) of accepting. RQL and Beta together define the “consumer’s point” of the oc curve.
RQL and Beta define the “consumer’s point” of the operating characteristic curve of the plan. (oc curve). It is called the “consumer’s point” because it satisfies the consumer’s intentions of usually rejecting lots if those lots are truly RQL fraction defective.
Some of the standards manuals neglect to mention RQL/LTPD/LQ. They bury the consumer’s point in the structure of the sampling plan tables so that you do not have to think about it. Using such standards, you cannot be sure that the sampling plan that you choose will classify a lot as acceptable or rejectable the way that you want it to. The advantage of such standards is that by using them you can say truthfully that you are following the standard. This might make you politically safe but not technically safe from a quality standpoint. More on this at:
www.samplingplans.com/aqlprimer.htm
The above description is for attribute sampling plans. If you have variables data you must additionally use the within-lot standard deviation. If your variables sampling plan is for the mean of a measured variable, AQL-mean and RQL-mean are used.
More on the design of sampling plans at:
www.samplingplans.com/modern3.htm