RE: Methods to Determine Special vs. Common Cause

[Jean-Marie Gogue <jmegogue@wanadoo.fr>]

On Sun, 6 Aug 2000 Vic Forte <vic@vichara.f9.co.uk> wrote:

>Does anyone have a definition of "special cause" using natural
>language. I don't think I have ever actually come across one.

Shewhart started his Theory of Variation in 1931 with a definition using
natural language, which, in fact, is an operational definition. Vic, I like
your question, because the Shewhart's definition is a cure against
superstitious beliefs concerning Theory of Variation and SPC.

According to Shewhart, a special cause of variation is a cause which is
likely assignable to something to be detected (a person, a procedure, an
event, etc.). The contrary of a special cause is a random cause, which is
defined by an operational definition.

A random cause of variation (synonym of common cause) is one of the many
causes resulting in a sequence of data which could have been produced by
the Shewhart's "Ideal Bowl Experiment". Usual criteria for special causes,
such as "one dot above 3 sigma..." have been derived from comparison
between an actual sequence and an "ideal" sequence. Now, what is the
"Ideal Bowl Experiment" ? Numbers between  -3 and +3 with ranges of 0.1 are
written on chips (about 1,000 chips). They are carefully mixed in a bowl. A
blindfolded person draws chips out of the bowl, one by one, and replaces
them in the bowl every time. Another one writes the results. The point is
to define randomness in operational terms, not in mathematical terms. It's
worth noting that the Deming's Red Bead Experiment is a simplified version
of the Ideal Bowl Experiment.

To summarize, perhaps we could say today that, by definition, a special
cause of variation is a cause which cannot exist in a situation like the
Red Beads Experiment. Do you agree ?

Best Wishes,

Jean-Marie Gogue
jmegogue@wanadoo.fr
========================================================================