From: Stan Hilliard
Date: 08 Jun 2000
Time: 01:43:41
Greetings Leonard,
The sample size should be determined by the oc curve. Consider what happens if an acceptance sampling plan of n=100, c=4 for N=2000 pcs were to be carried out as n=25, c=1, four times on partial lots of 500 pcs each.
If the results were accumulated over the four partials to make an accept/reject decision on the whole 2000 pcs, then the original oc curve would hold.
If accept/reject decisions were made for each of the four partial lots of 500 pcs instead of overall 2000 pcs, the original oc curve would not apply.
Here are the AQL and RQL of the two methods, using alpha=0.05 and beta=0.05 with program TP105.
1) Overall lot of 2000 pcs: AQL=0.0199, RQL=0.0892
2) Each partial of 500 pcs: AQL=0.0144, RQL=0.1761
Even though the overall amount of testing per piece is the same in these two cases, their decision schemes differ. It is the n & c that cause the difference, not the lot size. OC curves apply only to the populations (lots or partials) that are the target of the accept/reject decision.
Sincerely, Stan Hilliard