From: Stan Hilliard
Date: 8/10/2004
Time: 10:47:24 AM
ramesh,
When a lot passes an AQL acceptance sampling plan, for example, AQL=1%, passing does not determine that the lot contains equal to or less than 1% defective items. The true percentage of items defective in the lot can be 2%, 3%, 4%, 5%, or more. It can be MUCH more if the sample size is small.
It is a matter of probabilities. The probabilities are described by the OC-curve. www.samplingplans.com/usingoccurves.htm
To read the OC-curve, you must start with the horizontal axis. That is, "if the true lot fraction/percent defective = X, then the probability of acceptance =Y." Again, you can only go from X to Y, you cannot go from Y to X.
The reason that I said that there are two questions is that an AQL sampling plan can make a decision (based on n,Ac) but cannot estimate how high the percent defective might be. That estimate is made by the upper confidence limit.
The relationship between between the upper confidence limit and the OC-curve is that if the sample contains Ac defectives, then the upper confidence limit = RQL. (not AQL). The confidence coefficient is 100-Beta (in percentages.) If the sample contains less than Ac defectives, then the upper confidence limit is less than RQL.