RE: Complications

["Jonathan Siegel" <jmsiegel@umich.edu>]

Steve,

As I know you've said many times before, Deming and Shewhart made a
conscious decision to make their approach as simple as possible, with
the idea that it could be understood by ordinary line workers with high
school math. Deming wrote in one of his first books, Some Theory of
Sampling, that one should avoid using a technique too sophisticated to
be understood by those who use it, because the error generated by misuse
may be greater than that gained by the sophistication. Deming also
preferred techniques which permitted people to calculate things
themselves, rather than having to rely on computers, to permit a better
intuitive feel for the situation. Conceptual understanding is far more
important than precision.

By way of analogy, an analog clock gives some users a better feel for
understanding approximate time than a digital clock. Even though it is
less precise, it can be better at helping people learn how to understand
and manage time. 

If the people regularly using the charts are more sophisticated folks
with greater backgrounds, more sophisticated and precise techniques
might be appropriate. 

But if this means charting is left in the hands of specialists and
ordinary workers aren't the ones regularly using (and understanding) the
charts, than I suspect Deming might feel that one of the core purposes
of the control chart is being defeated. 

The issue of predictive ability of standard deviation vs. range involves
a loss function and also depends on purpose. 

Jonathan Siegel
734-994-8089
jmsiegel@yahoo.com

P.S.

I want to apologize to the Deming community for not contributing
recently. I've been in a statistics PhD program this past year and tried
to set all else aside to recover from setbacks and gain the mathematical
background needed to succeed. I've discovered I needn't have bothered.
I'm not a pure mathematician and will never be one.

One thing I often saw in the program was the application of tremendous
mathematical sophistication, often for a small gain in precision or
assumed reliability, and often based on adding questionable assumptions.
A favorite example I observed was a talk on a new method for conducting
agricultural experiments, of wonderful mathematical elegance, which
required the assumption that the terrain be absolutely flat. There
seemed to be a reward structure which rewarded complication and
sophistication for its own sake, rather than simplicity and utility.