RE: Info

[Rip Stauffer <rstauffer@bluefirepartners.com>]

Richard Bell said: "I know some schools are teaching this in undergrad
programs but I've got no clue how many or how widespread it is.  
"I suspect not as widespread as it needs to be."

In my Masters Thesis, I did a modest literature search. I only found one
article in the that discussed analytic and enumerative studies. Since then,
I know of at least one article that mentions it, in the Journal of Quality
Technology, but it only mentions it in passing. A majority of statistics
texts don't treat control charts at all; those that do often cite
dependencies on normality or other restrictions that don't actually exist in
practice. Others teach erroneous methods for calculating control limits,
e.g., using three times the standard deviation for the entire data set.
Rational subgrouping is ignored, Likert-scale data are averaged and used to
develop control limits. All wrong, all a potential recipe for waste or
disaster.

The difference between analytic and enumerative studies is not taught in
most statistics programs; most statistics courses are under the umbrella of
the school of mathematics. The logic is different, the focus is different,
and the assumptions are different; if you ignore these assumptions, your
conclusions will be questionable. This difference, however, is not taught.

George Box explained the problem as a difference in paradigms. "In the
1930's" he said, "an important issue...was whether statistics was to be
treated as a branch of Science or a branch of Mathematics." In his view, the
worst happened...statistics became a branch of Mathematics, so many
statisticians adopted the mathematical "theorem-proof" paradigm instead of
the iterative learning paradigm of the scientific method. In practice, then,
statisticians were trained to test solutions, but not to help discover
solutions.

Recently, I have seen some progress. At the Fall Technical conference a
couple of years ago, one of the primary presentations was around William
Woodall's paper in the October 2000 JQT, with some of the discussants (Roger
Hoerl and Don Wheeler). A lot of the discussion at the end of the
presentations centered around the apparent schism between academicians and
practicioners. Practicioners complained that their papers can't find peer
review and that academic papers often produce unnecessarily complex
solutions that don't work "in the field." Academicians complained that
practicioners often use techniques that are not mathematically sound and
have not been published or peer reviewed, and that some of the new solutions
are never tried in the field, so never evaluated. Learning opportunities are
being missed on all sides. 

The discussion ended on what I thought was a very positive note; the problem
seemed to be universally acknowledged and understood. Roger Hoerl pointed to
Technometrics as a journal that will publish more papers from practicioners.

Sorry for rambling. In my opinion, this is a very sticky wicket that has yet
to be unstuck. Myron Tribus once told me, in reference to this problem, that
he did not want to get involved in an argument over the difference between
analytic and enumerative studies. He thought "the only things that matter
are 'what question are you trying to answer?' and 'what's the best way to
get the answer?'" I, for one, think he's right; but it's no easy task
acquiring the wisdom to find the answers to those two questions.

Best regards to all,

Rip

Rip Stauffer, Senior Consultant
BlueFire Partners
1300 Fifth St. Towers, 150 So. Fifth St.
Minneapolis, MN 55402
612-344-1027
mailto:rstauffer@bluefirepartners.com
http://www.bluefirepartners.com/