Guest post by John Hunter, who founded curiouscat.com in 1996.
2015 ASA Deming Lecture by William Meeker – Reliability: The Other Dimension of Quality:
William Meeker begins the presentations discussing his experience with W. Edwards Deming in relation to a paper by Meeker and Gerald Hahn. The core issue related to the issues around problems in how analytic studies are treated. We have a previous post on this blog on this topic: Enumerative and Analytic Studies.
William also mentions Dr. Deming’s paper, On Probability As a Basis for Action, which we make available through our website. William summarizes what he took from the paper in reference to analytic studies:
- Subject-matter experts often needed
- Things change over time
- Interactions exist
- Future environment conditions will be unknown
- There is no statistical method to extrapolate
Dr. Meeker defined the studies this way, in the talk (and his paper with Gerald Hahn: Assumptions for Statistical Inference):
Enumerative study: Making inferences about a static population after sampling from a corresponding frame.
Analytic study: Everything else and especially sampling from a process.
Dr. Meeker showed a letter Gerald Hahn received from Dr. Deming on their paper:
Your paper is great. You refer to a new book that you published. I should be pleased to have a copy. To go on, your paper pleases me very much. In the first place, almost nobody seems to be interested in the difference between enumerative studies and statistical inference, which I called analytic problems. It pleases me much to see not only your interest in the distinction but great contribution thereto.
The book mentioned was Statistical Intervals (the link is to the 2nd Edition of the book).
Dr. Deming subsequently invited Meeker and Hahn to present their paper at an upcoming Deming seminar for statisticians. During their presentation there was a great deal of vigorous debating – much of it from Dr. Deming himself. I think this is an example of some of Dr. Deming’s strengths: his lifelong interest in continually learning (you can follow his lead and read the paper he spoke so highly of, which I linked to earlier in this post) and a willingness to explore new ideas even while challenging some of the ideas being presented.
The presentation continues on the topic of reliability and that section provides some examples of the challenges with analytic systems. As an example, cell phone relay device manufacturers not appreciating how damaging the conditions on cell towers would be which lead to more rapid wear and tear than they expected.
He also provides insight on gathering data during the use of a product to predict when it is wise to intervene. This doesn’t require plotting control chart data (sometimes that isn’t realistic – especially for products being used by customers). In several examples he discussed predicting the likelihood of failure based on factors during use and how collecting data (simple use data, etc.) can allow an organization to create processes to intervene prior to failure. In doing so they can wisely take action based on expectations of the failure likelihood become so high it is worthwhile to inspect, schedule preventative maintenance or proactively replace the old item with a new item.
We have now added blog posts on most ASA Deming lecture from 2009 through 2017 (and 2009), the others are:
- 2009, J. Stuart Hunter (Princeton University), New Challenges for Statisticians
- 2010, Sir William Osler: Dr. Deming Consults on Quality
- 2011, Roger Hoerl: Need Any Country be Poor?
- 2012, Dr. Jeff Wu: Quality Improvement from Autos to Nanotechnology
- 2013, Vijay Nair: Industrial Statistics – Research vs Practice
- 2014, Sharon Lohr: Red Beads and Profound Knowledge: Deming and Quality of Education
- 2016, Vincent Barabba: Profound Knowledge from a Knowledge Use Perspective (I will add a blog post on this one soon)
- 2017, Fritz Scheuren: W. Edwards Deming – A Kaizen Statistician