Downloaded from https://www.everyspec.com AMCP 706.198 usually optional, aid eith, r the horizontal If d cm + 1, reject the lot. Other- or vertical segment can be missing, wise continue sampling. Fig. 3-2(D) is an unusual sequentiai-sample N 1 = size of sample I plan; it has 3 decision regiuns on the bound- ary. For example, in the "worst," region, in maximum number of samples one might reward the supplier. In the supplier in some other way, In the "better" n = number tested so far region, one might reward the supplier. in the "OK" region, the lot would be accepted but d = number of defectives found so far with no penalty or reward.
i-sample Plan For each plan, the OC curve can be de- rived (in principle anyway, in practice it N, 100, c " 3 might be difficult, expensive, ,id/or tedious). It will show the probability of a particular This means that if it 100 and d 3, decision vs true quality. Also, the Average accept the lot, if d - 4, reject lot. Sample Number (ASN) can, in principle, be Often the simplest administrative Instcuctions derived for each plan, as a function of true are presumed to be: Test the whole lot of quality. 100; if "d , 3", accept the lot, otherwise reject it.
It is easy enough to estimate the actual quality of the lot, but it is quite difficult to 2-sample Plan find the probability distribution of that estimate unless the sample size is fixed. [n N, = 100, cn 3 all the plans shown in Fig. 3-2, the sample size is a random variable because a decision N, = 200, c:2 = 7 can (must) be made any time a boundary is touched (this is called curtailed-sampling). This neans that if n 100 and d ( 3, Since this chapter deals only with decision accept the lot. If n - 100 and 3 d 7, making, not estimation, the difficulties in take a sample of 200 more. If n 100 + estimating will not be considered further. 200 300, and d 7, accept the lot.
If at any time, d > 7, 1-eject the lot. As in the There is no rule or law of statistics (or I-sample plan, it is often thought to be anything else) that determines what shape simpler not to have the inspector concern the decision boundary must have. Some himself about the number of defectives, shapes may be better than others, according except at n - i and n N, + N 2 . to specific criteria, but no shape (or implied plan) is always better in all ways than any In the seq-sample plan, it is usually con- other plan. sidered that a dc,.l,,ion of some sort is made after testing each specimen.
In practice this Two examples are given for the usual need not be true, When d is discrete, it is m-sample plans, Notation follows: possible at the beginning to determine the minimum n (n = n, ) for acceptance (i.e., c w acceptance numbei, i.e., if d ej d = 0) and to take a sample of that size. when n 2; NI then accept the lot. If the point (n, , d) is within bounds, tie minimum "additional sample size to accept" .4~ 61