Twenty Years of Testing Questions
At a recent meeting of the Institute of Packaging Professionals’ Southern California chapter, guest speaker Frank De Meo, a staff statistician for a medical device company, tackled the sample size question. To determine the proper sample size for many statistical tests, one needs to know two quantities, he says: one, the accuracy of what you want to estimate, and two, the variability of the manufacturing process. “With these two quantities, there are relatively simple equations to determine the proper sample size. Simply spoken, once you know the required precision and the variation of your process, the sample size can be determined,” he reports.
Hearing such advice, I felt like someone had just handed me a calculator. It seemed to me that if you plug in your company’s standard deviation and your operation’s probability of defect, out should come your sample size for testing.
But I think this is where the challenges begin. Does every company have stated policies? Are all team members aware of them?
And how do you arrive at your defect rate? De Meo invited the group to participate in an exercise he called “the Blind Inspector” in which the defect rate changed over time. The good news is that “as the defect rate rises, the probability of acceptance goes down, in general,” he pointed out. But the trouble is, “if a given operation appears under control, there may appear to be no reason to look for a problem,” he warned.
De Meo advised attendees to follow ANSI’s Standards for Acceptance Sampling for ongoing production. Binomial distribution, confidence and reliability, and statistical tolerance levels are all essential details to understand. De Meo also advised instituting statistical sampling methods during receiving and inspection.
There are some other pitfalls. De Meo said other challenges are figuring out whether you are trying to get an estimate of your defect rate, and how accurate that estimate needs to be. Jan Gates, SoCal IoPP Co-Chair and with PackWise Consulting, says another question is whether testing is for validation or for on-going production. And Karen Greene, SoCal IoPP Co-chair and with Life Packaging Technology, said another challenge comes when you are working with unknown population sizes or huge populations.
There are many questions to address, and I may have just raised more. One thing is certain, though—there is a lot of support out there for you. I was thankful for this meeting, organized by Gates and Greene, and I encourage you to seek out future IoPP chapter meetings for ongoing support.
In fact, Greene will be speaking in an upcoming IoPP webinar on October 16 on this very topic: "Determining Appropriate Sample Size for Package Testing." According to IoPP, Greene will cover the following:
- Understanding compliance and guidance documents for sterile barrier systems.
- Statistics reference books and articles.
- How to determine sample size for final packaging system validation.
- Tips for developing an appropriate and statistical valid test population for SBS validation.
- A summary of anonymous industry interviews on sample size for sterile barrier system testing (not a statistically valid sample size).
For more details, please visit IoPP's site.