From: Stan Hilliard
Date: 30 Jul 2002
Time: 00:21:38
Greetings Erik,
You need to classify each of the characteristics as to whether it are attribute data (each item is go/no-go) or variables data (numerical measurements). Standards are classified that way.
The typical standards are ANSI/ASQC 1.4 for attributes and ANSI/ASQC 1.9 for variables. Both are available from ASQ. Both of those standards are based on AQL percent if defective items.
You should probably have a copy of those standards because they are the most used. However, I think that they are insufficient for developing a sampling scheme which you can properly understand. A shortcoming of Z1.4 and Z1.9 is that they are based on AQL. The AQL method is not a risk/consequence based method.
See www.samplingplans.com/aqlprimer.htm
If you want to really analyze your situation closely you should use the Operating Characteristic (OC) curve method. It is the only way that you can balance the cost savings and time savings against the risks involved with sampling.
See www.samplingplans.com/usingoccurves.htm
The software programs TP105 (for attribute data) and TP414 (for variables data) will do the calculations for the oc curve approach to sampling. The programs use the oc curve to develop fixed sample size and sequential sampling plans.
www.samplingplans.com/software_oc.htm