RE: Rule of thumb for SPC

[Steven_S_Prevette@RL.gov]

QUESTION:
One of the things I've noticed in working with control charts over the
years
is that when you delete out-of-control data points, frequently the
control
limits "shrink" leaving you with more points that are now outside of the
newly set control limits. Has anyone else encountered this and if so,
what
are your thoughts regarding this phenomena? Are these points truly a
signal
of non-random variation in the process? 

RESPONSE:

Yes, I have seen this phenomenon.  Definitely these are signals that the
process data are not stable.  Usually, if after throwing out one datum
point and the narrowed limits cause another (well separated in time from
the first) to fall outside, I will throw the second out.  If that causes
yet another individual to go out, I give up and put all back in.

I must point out that one should go looking for the special causes of
these points.  I must admit that many times the analyst does not know
what special causes there might be when they receive the data, and must
(OK, maybe "must" is too strong a word, but even analysts have
customers) make some judgement on setting baseline averages and control
limits without full knowledge of the process.  In the case of recursive
recalculations throwing more points outside the control limits, this
becomes a case where knowledge of the process is an absolute necessity,
which is why I give up after the second iteration.

Steve Prevette
FDH ESH&Q Planning and Evaluation
ASQ Certified Quality Engineer
======================================================================