From: Stan Hilliard
Date: 1/24/2007
Time: 2:20:47 PM
"Statistically valid" means that you know and agree with the probabilities of acceptance of various quality levels -- as reflected by the oc-curve.
samplingplans.com/usingoccurves.htm
Two companies can have grossly different sampling plans and each would be valid if the plans behave the way they each intend.
A sampling plan can be selected from a standard like 105D or Z1.4 and still NOT be statistically valid if its performance is different from their belief.
The sampling program TP105 designs sampling plans for attribute (defective/not-defective) data.
The sampling plan program TP414 designs plans for measured data. The kinds of plans that TP414 designs are for:
1) Measured individual items with specification limits that apply to individual items (ISLs). This type of plan is referred to as sampling for variables for percent nonconforming.
2) Sampling plans for the lot mean. This type of plan is appropriate to material where there are no individual parts -- like bulk liquids, powders, color measurements, etc.
All of these plans use the oc-curve to reflect their probabilistic performance. The oc curves are indexed by their producer's and consumer's points.
samplingplans.com/software_oc.htm