Re: The rationale for substantive knowledge of underlying process

[GrantBlair@aol.com]

In a message dated 5/9/2001 4:07:47 AM Eastern Daylight Time, 
jsdwd@ksc.th.com writes:

<< I'd disagree to the extent that in order to plan for rational sub-grouping
 some process knowledge is necessary.   One needs to understand enough to
 know how various sub-grouping possibilities would be implemented to put
 together a sensible set of limits for a given process.
 
 Note that the determination of special v. common causes is an operational
 function and that the control chart represents and operation that enables
 one to make this (essentially) economic distinction. >>

John makes two significant points:
 Selection of the rational sub-group is extremely important. For example, I 
could call for 4 repeated measures of the same piece (Maybe even the same 
location on the piece!), and call this a rational subgroup. Although this 
would be rational for
measurement error, it would be irrational for process control. Typically, 1/3 
(or more) of the points would indicate special cause. My personal preference 
is to select a rational subgroup which produces approximately 2 percent, ( 5 
percent, tops!) special cause signals.....using 3 sigma limits and 
appropriate runs rules.

This leads me to the second point, which is the economic balance between 
special and common cause. I state this as an operational definition..."The 
PURPOSE of a control chart is to separate common cause and special cause 
variation.
Unless this is done, any attempt to IMPROVE the process will only result in 
making it worse". Note the emphasis on process improvement. Two per cent of 
the points can usually be economically investigated by the workforce to gain 
process understanding. Although the probabilities of false signals may be 
similar, (but lower), experience has shown that the collection of 
investigations over time, (both successful and unsuccesful) can be analyzed 
using the Pareto principle in order to enhance the workforce's process 
knowledge, and provide direction for Management's efforts to
continuously improve the process. 
However, if there is no interest in either of these, then a rational subgroup 
can also be selected to produce control charts
indicating a stable process, as Kromkowski suggests.

GrantBlair@aol.com
Ninety Six SC