Guest post by John Hunter, author of the Curious Cat Management Improvement Blog (since 2004).
George Box shared a presentation on Rethinking Statistics for Quality Control at our 2008 Deming Institute Conference in Madison, Wisconsin.
In the presentation George discusses how to look at data from a process. He mentions why it was so important to understand what Shewhart understood about process data: the order of the data is extremely important; which is why run charts and control (process behavior) charts are plotted in time order.
The talk captures George’s humor as did what I think is a very entertaining and interesting autobiography: An Accidental Statistician.
At points the talk does get into statistics beyond what many of those reading this blog are likely to need, but if that is true for you it is easy enough to ignore those sections. I liked George’s comment during one of those forays:
Look George, I know how to calculate an average, but this is all very complicated and I don’t like it much.
My guess is if you are feeling that way when you reach this point in the talk you might want to skip the next 10 minutes of the presentation. I find it interesting but it is getting into statistics that is far beyond what is normally needed. For complex processes (large chemical plants, complex manufacturing processes…) these ideas are important but most of us will not need to get into these complex statistical engineering ideas.
The discussions at the end are interesting. George talks of adjustments (in the talk and the questions and answers) which I can imagine many with an understanding of Deming’s ideas would think of as tampering. I do see that risk in the actions he discusses. But I also think we should be open to thinking about alternative strategies; not to just ignore the risks of tampering, but consider alternatives with understanding of those risk while looking to most effectively manage existing processes.
Then, with that understanding, evaluate the results of the adjustments. Do they result in the mess tampering can lead to, if so, stop. Does it result in benefits that are measurable and as George mentions several times in the talk, robust, if so that is likely a useful strategy in the context it is being used. George wrote a book on the topic, with Alberto Luceño: Statistical Control: By Monitoring and Feedback Adjustment.
In case you are wondering if that is Gipsie Ranney asking some of the questions and making some of the comments, it is. You will also hear Ian Bradbury asking questions (it is often difficult to hear the questions, I realize).
We have also posted Peter Scholtes’ presentation from this same conference online: Deming 101 – An Introduction to Dr. Deming’s Teachings.
Related: George Box Webcast on Statistical Design in Quality Improvement – Articles by George Box – Quality Comes to City Hall – The Art of Discovery – Exploring Measurement, Presentation by Ian Bradbury