RE: Shewhart's Maximum Control

["Mowery, R. Neal (RNM) " <RNM@y12.doe.gov>]

	Lloyd Provost wrote:

	>Neal, are you familiar with any research on the ability of
individual value
> control charts for skewed data (like cycle time) to detect special causes?
> Don
> Wheeler has done a lot of work on the affect of various distributions when
> the
> system is stable, but I am not familiar with any papers that describe how
> the
> skewness affects the X-charts ability to detect special causes.  Some of
> my
> practical experiences indicate that an X-chart established for cycle times
> without a transformation is not very sensitive when special causes occur.<
> 
	Lloyd, other than Wheeler's writing (the Empirical Rule in
"Understanding SPC") on XmR charts for non-normal distributions, I am not.
I guess this points out the need for independent research, though, because
my experience has been just the opposite of yours.  I find that if I use a
transformation of, say, a lognormal distribution, I end up with a UCL that
is so extreme that it becomes very IN-sensitive.  I use the moving range to
calculate the UCL for X.  Maybe that is a difference.

	I would defer to others who have spent a lot of time looking
specifically at cycle-time however.  I have worked with similar
distributions for strength, contamination, flatness, gap size, etc., and
have mentally grouped them into a class where variations (errors) in process
tend to move the result in a single direction (toward bad: e.g. weak, dirty,
warped, large gaps, or in cycle time, slow ).  Perhaps there are nuances I
don't see in cycle time.  But it has worked for me when I have used it.

	I wonder what others have seen?

	Neal


*********************************************
Neal Mowery 
Statistician - Lockheed Martin Energy Systems
Oak Ridge, Tennessee
moweryrn @ y12.doe.gov     Voice (865) 574-0796
*********************************************
==================================================================