From: Stan Hilliard
Date: 01 Jul 2003
Time: 17:09:33
Russel said: "My question is can I have an attributes plan for this test wherein a bottle falling within the limits will be deemed non defective and a bottle falling out of the range as defective? Which type of plan would be better here, variables or attributes?"
You can produce variables and attributes plans that are matched -- that is they have the same oc curve. Thus they will perform equally with respect to accepting and rejecting. Matched plans have equal ability to discriminate between AQL and RQL lots. See:
samplingplans.com/modern3.htm
The main difference is that the attribute data is less efficient than the a matched variables plan, and thus the attribute plan will require a larger sample size.
Which is best is a matter of economics for a particular application -- considering the cost of the larger sample size in time and materials and the difference in the ease of collecting the two kinds of data.
If there is a need to use the data for problem-solving later, variables data would have the advantage.