From: Stan Hilliard
Date: 11/4/2005
Time: 10:02:11 AM
Hi Stan,
"... Can someone fax that article to me...". It is done.
The article explains why the AQL based C=0 plans do not offer the protection that most people think they do.
I have tried to clear up the misconception about C=0 with a tool for C=0 plans based on RQL instead of AQL. I added it to the new windows version 3 of TP105.
The calculated sample sizes increase significantly when C=0 plans are based on RQL, but that avoids the misconception of false sense of protection. The RQL based plan is statistically identical to the plan that accountants use called "Discovery Sampling".
The sample sizes are high with the RQL based plans, but it is possible to reduce the sample size drastically if the quality characteristic is a measured variable. (The sample size was reduced a factor of 25 in one example.) The oc-curve of any RQL based attribute plan can be matched with a variables plan having identical producer's point and consumers' point. TP105 v3 performs that matching to SD unknown and SD known variables plans for percent nonconforming. ISL specification limits can be lower, upper, or both.