The concept of Statistical Thinking has always appealed to me since I first heard it back in the late 1970s or early 1980s. The concept surfaced, in my mind, recently when I began teaching two sections of Elementary Statistics at a local college in Northern Illinois. I told my students that developing a sense and adopting the concept of Statistical Thinking is crucial, in my opinion, as mastering the definitions, formulae, and computations we would be covering in the course.
I knew this was the right thing to say. I was proud of myself. Then I was relieved that no one actually asked me what I meant by Statistical Thinking? I would have stumbled around for an answer. I intrinsically knew what I meant but I did not have a good definition ready to deliver and expound on.
What did I mean by Statistical Thinking anyway?
I meant, in so many words, the ability to NOT react to the world as if everything were solid and unvarying. There is variation in everything around us. Yet, we should not act and react on small changes that are part of “normal” variation but rather act and react when the variation or trends are “significant” enough to indicate a real change in the state of things. This is generally correct, but hardly concise, easy to understand, or compelling. I would have confused the students more with my awkward, wordy, and meandering explanation.
Beyond the classroom, this concept of Statistical Thinking would be very helpful in our company, DemandCaster. We are constantly looking at historical demand to prepare forecasts. We want to look and account for special causes of variation while being mindful not to overly react to the natural or common cause variation in these time series. So, I needed a good short concise definition of Statistical Thinking.
So, what to do? Where could I get some assistance? There were many options. First, I googled it. I didn’t get anything compelling, probably because I did not really dig very deep. I decided to post a question on three American Society for Quality Groups that I belong to on LinkedIn. I found this to be a good idea as I received many interesting responses including “Why don’t you google it?” Duh.
If there is an official definition, here is one that is probably it:
Statistical Thinking is a philosophy of learning and action based on the following fundamental principles:
1. All work occurs in a system of interconnected processes
2. Variation exists in all processes
3. Understanding and reducing variation are keys to success.
This was posted by John Surak, Cliff Lee, and others. This is taken from The Glossary of Statistical Terms, ASQ Quality Press and attributed to a paper by Lynne Hare, Roger Hoerl, and Ron Snee who were giving a paper titled Statistical Thinking for Business Improvement at the 49th Quality Congress.
Here are some of the other responses from LinkedIn Group members on their views of Statistical Thinking.
• Statistical Thinking is the ability to understand the sources of variability inherent to all processes. – Cicero de Melo Lucas
• H G Wells (1866 – 1946): Statistical thinking will one day be as necessary for efficient citizenship as the ability to read and write. – Contributed by David Ehlers
• Michael J. Armstrong contributed some gems:
– Management is the art of making sufficient decisions given insufficient information.
– Statistical thinking is the art of making deterministic decisions given probabilistic information.
– Statistical thinking is how one can achieve a consistent understanding of an inconsistent process (and still remain sane).
• Tonya Connariato pointed out that “ASQ has a blogger who's writing specifically on this topic at http://community.asq.org/statistics”
• John O’Mara cracked me up with this one: I'm not familiar with the term from the 1980s but what came to mind for me was what Monte Visser once told me, "You know that movie where the kid keeps saying, ‘I see dead people?’ Well, I see distributions."
Having read all these comments, I decided to see if Kaoru Ishikawa had anything to contribute to this definition. For me, Ishikawa always nails it. So, I perused a book that was, another big duh, on a bookshelf right next to my desk. I found Ishikawa, Kaoru, Introduction to Quality Control, 3A Corporation, Tokyo, Japan, October, 1990
The Statistical Approach
To start, we should understand the following four things about the statistical approach:
1. Results of any work we do always contain variation and follows a certain distribution pattern.
2. Error is a basic concept; the data produced by business enterprises and organizations in society include dirty data, abnormal values, and false data.
3. Data are always collected with an intention to take action.
4. Stratification is another basic concept; everything should be thought of in a stratified way, and everything must be stratified (i.e. segregated into meaningful streams or groups) for data collection and analysis.
So, what did I learn? Well, I certainly could not argue with the ASQ responses or Ishikawa definition. They all really captured the concept quite well. However, these definitions are used for people who want to learn to act and react to variation in an analytic way. I envision these definitions being used in the first few slides of an Introduction to Quality Control course. In fact, I am sure that is exactly where these definitions are most used.
I was still looking for something a bit more pedestrian, conceptual, something I could use in explaining the concept to someone, well like Mother, who might find the ASQ or Ishikawa definition a bit confusing.
What is my elevator speech on Statistical Thinking? Here is my humble attempt which I am considering a work in progress.
Statistical Thinking is a way of looking at everyday and business information and being aware of the variation inherent in the numbers. It is also an awareness of NOT reacting to the “normal” fluctuations in the data but rather to look for significant changes that would indicate that something has changed. Statistical Thinking is a mindset to account for this variation or uncertainty as we make decisions every day.
I might certainly get questions back asking me what I meant by “normal” fluctuations versus significant changes. I will work on that answer next.
