Complications

[Steven_S_Prevette@rl.gov]

Charlie asks "  How complicated will we make this for people without your
backgrounds," with respect to the statistical discussion on moving range
versus standard deviation.

I should point out that Dr. Deming did state that statistical leadership is
necessary for a company.  Beware of hacks.  You need someone with knowledge
of statistical theory.  Refer to chapter 16 of Out of the Crisis for further
discussion.

Charlie, I see you are with the Navy.  The Navy educates at least 150
officers per year at the Naval Postgraduate School in statistical
techniques, at a Masters level.  Theoretically, the PG school staff should
also be there as a resource.  I cannot speak for PG school policy (as far as
how to get help, how much it costs), but I can state my thesis adviser is
now head of the Operations Research department, so I could put you in touch
with him.  I assume that the school still sends students out on 6 week
experience tours (granted, you do have to pay their per diem), and students
are also looking for thesis work.

For a "small company", I would agree that they may not be able to employ a
full time statistician.  But in many cases there may be an engineer or
accountant with sufficient mathematical background who could be trained and
mentored.  The American Society for Quality should be a resource in this
manner.  I try to do what I can through these DEN postings.  There are many
consultants out there who would be happy to help (for a fee, usually, but
there are many pro-bono ways to get advice).

In the case of the "small company" without a full time statistician, I would
recommend sticking to the tried and true, and stick to one expert's
methodology.  Certainly Dr. Wheeler's series of books are a good, internally
consistent model to follow.  In the matter of the range versus sigma, I
would stick to the traditional range method.  Yes, I am using the sigma
method, but as can be seen in postings on the DEN, it is somewhat
controversial.  It is not worth letting the controversy interfere with a
basic decision as to whether or not to do SPC.

The benefit of SPC is getting companies to move away from "everything is
special cause".  Generally, the vast amount of our problems are common
cause.  I would actually trade off the loss of detecting some special cause
in order to get managers to look more at common cause.  

We can have academic discussions of what version of SPC will detect special
cause variation the fastest.  But a key is to pick a consistent methodology
(don't change methodologies in midstream in order to get a more "favorable"
result) and execute it.  Go for feasibility first, optimize later.  Follow a
known methodology.  True, bad habits adopted early on are hard to undo
later, but don't let that prevent you from taking action.



Steve Prevette
Site Technical Authority for Statistical Trending
Environment, Safety and Health
Fluor Hanford, A Fluor Global Services Company
ASQ Certified Quality Engineer
steven_s_prevette@rl.gov
509-373-9371