My Sigma's bigger than your's

[John <jsdwd@samart.co.th>]

on 6/10/00 12:23 am, GrantBlair@aol.com at GrantBlair@aol.com wrote:

> In a message dated 10/5/2000 8:31:55 AM Eastern Daylight Time,
> jsdwd@samart.co.th writes:
> 
> << The statement above that until special causes are removed 'any attempt to
> reduce variation ...will only result in making it worse.' is simply not
> true. >>
> 
> Here is the WHOLE sentence from my post:
> "The purpose of a control chart is to seperate common cause from special
> cause variation. Until this is done, any attempt to reduce variation in a
> process will only result in making it worse."

I fail to see how adding the additional verbiage changes the meaning one
whit.  That's why I left it out.
> 
> Here is what I also said in a earlier post responding to you:
> 
> "Control charts define WHEN to take action on the process. In their absence,
> variation will increase:
> First, by taking action on a process when it is not required.
> Secondly, by failing to take action on a process when it is required."

I will make my point yet once again.  Let's consider a simple example of a
process operation that uses fixturing of some sort; perhaps a fixture
holding another part that is being ground or cut...use a little imagination,
I'm sure you can see such a case easily.  A design engineer has been working
on a new method of fixturing that will hold the part more firmly giving a
more uniform cut with smoother edges.  Let us further assume that smooth
edges are a quality characteristic of interest.

This factory has come late to the control charting method and is only now
beginning to control chart some of their key quality characterisitcs such as
smoothness and the initial control charts show an abundance of special
causes (as is often the case....NOT ALWAYS!).  Simultaneously the fixture is
shown on the basis of some trial runs to markedly reduce variation in
smoothness.

Would it be your position Mr. Blair that the new fixture should not be
introduced to the process until all the other special causes were
eliminated?  Would it further be your position that introducing the fixture
at this juncture would increase variation?

I think this rather simple argument shows the flaw in your statement above.
I hope that you and others who may be confused by this exchange will see
more clearly what I am trying to say.  

A more correct statement by you might be, "Control charts are one method
that can be used to define when to take action on the process.  In their
absence it is not uncommon for process variation to increase.  Until special
causes are identified, it is quite possible that taking action on a process
may make variation worse."

I wouldn't disagree with that statement.  Perhaps I'm just being picky, but
language is how we communicate in this forum and it is, in my judgement,
important that we take care to say what we mean....Deming was very careful
with language and I have learned to try to be careful as well, although I am
clearly not always effective in communicating my meaning.


> Read the WHOLE book..got it sitting here in front of me.
> Seen Don Wheeler's WHOLE video.

While I admire your obvious feeling of accomplishment in having completed
the entire book and the entire Wheeler video, may I respectfully suggest
that you read the book again? (I've read it several times, but I may be
particularly dense).  In any case, you seem to have missed some of it.  let
me suggest particularly the section beginning on page 39, entitled 'Two
kinds of errors in the operation of control' - Dover Press Edition. (If you
have the publication put out by the Government Department of Agriculture I
can cross reference for you.)

This is  where Shewhart contrasts the testing of hypothesis and control
charting and de-bunks (therefore) probability limits and some of the earlier
postings (yours among them) about tolerance limits, predictions, etc.  Also
on page 69 of that same book, Shewhart (discussing the setting of tolerances
which we debated earlier) says, "Whereas there is safety in numbers on the
basis of a sample from a bowl, that same degree of safety does not exist
when the samples are not so drawn."  In other words, the limits are
empirically derived...a point Wheeler makes quite often.

> 
> [Moderator's Note:  I hope that both 'sides' of this 6 sigma/SPC discussion
> will remain professional and not allow frustration to make it personal. ]

While I do appreciate the moderator's concern, and gentle chiding.  I have
resisted the urge to abandon the kind of tit for tat nonsense that has
accompanied some of these posts because I think there is something to be
gained in clarifying some of these issues.
.
There are at least two kinds of ways one can contribute here. One is to
present one's own ideas for consideration and the other is to (more or less)
publicly disagree with those who put forth ideas with which one does not
agree.  I don't see any problem with either method, but as I said in my
earlier post, why be pugnacious about it?  It blocks learning, which is what
I thought we all were involved for.

Also, I do think that one of the purposes of this forum is to make other
readers/participants aware that the force or assuredness with which an
argument is presented does not affect it's merits. As Dr. Deming used to
stay, "I have seen very gifted presenters teaching spellbound audiences
substance that is totally wrong." (or words to that effect).

In any case after a 20 year consulting career, I've developed a pretty thick
skin...I won't melt.

I also don't see what this discussion has to do with 6 sigma, in any case.

Now that's a WHOLE other kettle of fish....