From: Stan Hilliard
Date: 9/21/2004
Time: 8:36:26 PM
Ronaldo, with destructive testing, I assume that low sample size is your aim. Here are a few ideas:
1) Use variables sampling plans where possible -- as opposed to attributes plans. Variables plans require much smaller sample size to attain the same producer's risk and consumer's risk.
2) Use sequential decision rules as opposed to fixed-n rule where possible. Sequential plans use much smaller sample size than fixed-n plans and work for both attributes and variables data.
3) Minimize the within-lot variability of the packaging materials if possible. This will reduce the sample size of a variables plan.
4) Control the lot-average of the quality characteristic close to its target. This will greatly reduce the sample size of a variables sequential decision rule.
(3) and (4) may involve helping your vendors to control their manufacturing processes.
5) When defining a "lot" of packaging material, use the concept of rational subgroups. That is, don't let different production runs be combined into the same lot. The same goes for any other factor that could increase the variability of the characteristic within a lot.
These concepts are discussed further in the tutorial at:
www.samplingplans.com/modern3.htm