Valid Statistical Techniques Made Easy

Course description

Why do people fear statistics? Could it be:

  • The subject is too complex and difficult to understand?
  • They get overwhelmed by the terminology in regulations, standards, and guidance documents?
  • Statistical software tools are a maze of terms and menu paths?

Good reasons to be fearful. We are providing an avenue to cut through the jungle and find safe pathways to understanding and immediate application. We will give many examples of application by medical device and pharmaceutical manufacturers.

This 3-day, computer-based workshop provides product and process experts in regulated firms a solid understanding of how to use valid statistical techniques for their needs. It focuses on specific methods and tools that assist with the challenges of Process Analytical Technology (PAT), Design Control, Process Validation, CAPA, and other mandated activities. The content is based on the requirements of the US FDA’s Quality System Regulation, ISO 13485, and the recommendations of the Global Harmonization Task Force (GHTF).

Industry-specific examples are provided throughout the workshop. The statistical material is covered in a practical manner. It is not over-burdened with statistical complexity. Guidelines, rules of thumb, and a very user-friendly software tool are given to foster correct, immediate, and sustained application.

Course outline

  • Fundamental Concepts for Quality Systems
  • Performing Process Capability Studies
  • Conducting Test Method Validation (aka, MSA and Gage R&R)
  • Sampling and Drawing Conclusions
  • Creating Acceptance Sampling Plans
  • Applying Control Charts

The primary goal is for all participants to be able to immediately apply valid statistical techniques. Students are provided a Participant’s Guide and very user-friend software. The Guide is filled with examples of application. The participants will receive data files so they can work in teams to solve practical problems. Thus, the workshop stresses practical application instead of theory.

Case studies

Who should attend

The content should have great value to industry professionals involved with: research and development, new product development, process development, manufacturing engineering, operations, quality engineering, quality assurance, supplier quality, and regulatory compliance.


There are no prerequisites for this webinar. No background in statistics is necessary. That is the foundation of our “keep-it-simple” approach. A basic familiarity with Microsoft Excel would be helpful, but not required. Participants will receive Microsoft Excel-based software and Excel data files. The software requires a Windows operating system and Excel software. The software and data files must be loaded prior to the workshop.

Equipment and software needed

A statistical software package will be provided. It is a Microsoft Excel application (SPC XL)
This software is only compatible with Windows operating systems.
System requirements:

Case study examples



I keep hearing that we only need three production runs for Performance Qualification (PQ). Our lead Quality Engineer keeps telling me that this is wrong. Who is right?

Scenario: For a new production process we will have two identical production lines and intend to run two, 8-hour shifts. Our process engineers insist that we only have to measure the results of three batches to satisfy the FDA for process validation. They say: “We’ve always done it that way. Why should we change our methodology? “Our lead Quality Engineer claims that there is no valid statistical justification for three batches. It looks like we are in for a “bad stat day”…

Key Question: Is there a valid statistical justification for the “three batch rule”? If not, what is a valid statistical rationale that everyone can understand for our PQ protocol?

Solution: Is provided in this course!



A competitor just received a Warning Letter from the FDA for misapplying standards for sampling procedures for inspection by attribute. We are in an acronym jungle: AQL, AOQL, LTPD, LQ, CRQ, Ac, Re, α, β, … What is important for us to do?

Scenario: You are a Quality Engineer concerned with sampling plans. Several standards are given to you by the senior QE. You cannot understand them, especially the ones related to inspection by attribute. Worst yet, an internet search found an FDA Warning Letter concerning the misapplication of a related standard. After reading it, you get the feeling that the other company did not focus on risk to the end user.

Key Question: How do you use standards or valid statistical techniques to focus on risk to the end user? Is there a straight-forward procedure to follow?

Solution: Is provided in this course!



For a new device we will only manufacture one device every month. How can we satisfy the FDA’s requirements for Performance Qualification (PQ)?

Scenario: We manufacture a complex, expensive scanning device. For a new model we anticipate only making one device every month. There are several critical measurements that we take before we decide that the device is acceptable. The FDA requires that we show that we can make consistent product over time. Since our production rates are so low, we feel that we cannot apply a valid statistical technique.

Key Question: Should we document that statistical techniques are not appropriate for this situation? If not, what valid statistical technique can we define, document, and implement?

Solution: Is provided in this course!



The customer of our components is finding more defective items than our inspection system finds. Is there a simple way to find agreement between our inspection system and theirs?

Scenario: Your company is a supplier of components to a medical device manufacturer. These components have very tight tolerances. Your company purchased several different types of expensive measurement equipment. The CMMs purchased are highly automated. Your processes have been validated and you certify the quality of everything that you ship to your customer. Unfortunately, your customer’s Quality Control group frequently rejects items for being out of specification. Your management wants a simple and inexpensive solution to this very bad situation.

Key Question: Is there a simple, inexpensive solution that is statistically valid? Should that be applied to automated measurement systems as well?

Solution: Is provided in this course!



We perform safety-critical attribute inspections of our device but are only 90% effective. How can we improve our effectiveness?

Scenario: We manufacture an implantable, electronic device. The device has a thin metal casing that is sterilized. We laser-etch the serial number of each device onto its casing. Sometimes the laser energy is too high and the etching cuts through the thin casing. If this happens, then the device cannot be sterilized. This is a safety-critical issue so our goal is to detect 99.9% of all defective casings. We tested our three inspectors and found that each one only detects around 90% of the defective casings. An automated inspection device can be built but will be extremely expensive.

Key Question: Should we buy the automated inspection system? Should we fire these inspectors and find some who are better? Is there an inspection methodology that will assure us of 99.9% effectiveness?

Solution: Is provided in this course!



The FDA Quality System Regulation says that we should apply appropriate statistical techniques to detect recurring quality problems. Is there an easy way to meet this requirement? What if our quality problems are rather rare?

Scenario: A company manufactures hundreds of different combination devices. An internal audit found that certain similar production steps are sometimes missed or performed incorrectly. These can affect product quality or production-line safety. The audit also found that while CAPAs were initiated, there was no trend analysis to see if these issues were recurring. Management reviews that are conducted every three months were ineffective in detecting trends. Both process and quality engineers claimed that these events were relatively rare and weekly, monthly, or quarterly trending would prove to be non-productive.

Key Question: Is there a valid statistical methodology that can be employed to trend relatively rare events? Will that help managers be more pro-active and also satisfy US FDA quality system requirements?

Solution: Is provided in this course!



How do I qualify my inspection, measuring and test equipment to satisfy FDA requirements? Do I have to do this for automated test systems?

Scenario: Your company just purchased expensive, automated measurement equipment, called CMMs. After Installation Qualification (IQ) the supplier of the equipment claims that the CMMs are qualified and ready to be used. Your Quality Engineer (QE) says that this is not enough. She is concerned that there are more considerations beyond installation.

Key Question: What else should be considered beyond installation of measurement equipment? Must that be applied to automated measurement systems as well?

Solution: Is provided in this course!



IndustryMedical Device, Pharmaceuticals
LevelAdvanced, Intermediate

Peter L. Knepell