Re: Statistical Control and Purpose

[GrantBlair@aol.com]

In a message dated 1/28/2000 4:02:50 AM Eastern Standard Time, 
Steven_S_Prevette@RL.gov writes:

<< I can only answer with my personal observations with interventions into
 processes that appeared to have a low level of management control, purpose,
 or aim.  I found most of them had already achieved a state of statistical
 control.  The theory I have proposed to myself is that given a long time
 interval for things to settle out, and with processes that have a low level
 of external factors impinging upon them, a form of predicatibility settles
 in. >>

This ground was covered in an earlier discussion. It is a fundamental law of
nature that all systems are inherently unstable and will tend toward 
instability
if left unattended. This is known as the law of entropy.
There are some consequences of this principle:
1. The act of measuring a system does not confer stability  (unless you are 
measuring subatomic particles <grin>). Measurement should be used initally to 
gain an understanding of how the system works, secondarily to stabilize the 
system, then finally to achieve improvement through reduction of variation.
2. Measurements which are not meaningful will appear to be stable. A friend 
of mine learned this the hard
way. While being trained on a process he was told that it was stable as long 
as the chart pens stayed within certain limits.
Months later, when it was apparent the process was unstable and the chart 
still looked o.k., he asked an electrician
to replace one of the sensors. He was amazed to learn the sensors were not 
connected to the chart!!!
3. Control charts which appear stable should also be viewed with suspicion. A 
good example is a chart
    with nearly all  points between +/- 1 standard deviation. This usually 
happens when the wrong rational subgroup
    is selected, (which is a topic beyond the scope of this post.) A good 
rule of thumb for a control chart is
    that there should be about 1-2 points/100 points outside of control 
limits. If this is not the case, then look
    at changing the sampling method until this goal is attained. Then, 
investigation of these points over time should
    lead to process understanding and improvement.

GrantBlair@aol.com
Ninety Six, SC
==========================================================================