[EMAIL PROTECTED] (Robert J. MacG. Dawson) wrote in
news:[EMAIL PROTECTED]: 

>      Isn't there a QC teaching expreriment (by Deming?) designed to
>      make 
> this point - that the strategy of chasing your last error is not as
> good as a consistent one?

Yeah, it's known as the funnel experiment; it involves dropping marbles 
through a funnel onto a piece of paper and recording where they hit.  The 
objective is to minimize the spread of the hit points.  It shows that 
various strategies for moving the funnel based on how far off the "target" 
the last marble hit all result in greater spread than keeping the funnel 
stationary.

The point is that you simply can't cancel out *random* variation with 
negative feedback; if a process is varying excessively and much of the 
variation is random noise, the only way to stabilize it is to locate and 
remove the noise sources, not to apply a classical control system to the 
process.
.
.
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at:
.                  http://jse.stat.ncsu.edu/                    .
=================================================================

Reply via email to