SPC Theory and Practice

["John McConnell" <wysowl@msn.com.au>]

 There has been a lot of discussion on SPC theory and the use of x and mR
charts in lieu of attributes charts recently.  If this subject is of
interest to you, why not try doing some experiments?  (Especially if like me
you have little or no formal training in statistics and are trying to
learn.)  Take some Red Beads and plot these as an attributes control chart,
using the formula for an pn chart.  Then use the same data in an x and mR
chart.  Are they different?  Why?  By how much are they different?  In what
respect?

Then find some stable Poisson data (perhaps from a text) and repeat the
experiment.  

Then go back to all four charts and alter a few of the data to deliberately
introduce a couple of Special causes (perhaps by adding or subtracting about
1.5 to 5 sigma to each data value selected).  Re-calculate the limits.

Finally, for about the last half of each data set, add about 1.5 sigma to
each data value to create a "shift" up in the centreline and re-plot the
points and recalculate limits.

Then go back to your text books.  They might make a little more sense.  They
did for me.  An important learning for me was that no two writers of
statistical texts agree completely.  The same can be said for any
discipline.  Perhaps the trick is to try and understand where they differ
and to try to determine whether the authors differ in aims, and/or
knowledge, and/or values and beliefs (this feels like a good time to give
Bill Scherkenbach's book "Deming's Road to Continual Improvement" a plug).

Then try the same experiments with some data from work.  Hopefully, new
light will have been shed on these data as well as the different approaches.
In particular, try plotting some data you are pretty sure are Poisson or
Binomially distributed (as well as non-normal) using both an attributes and
an x and mR chart.

Deming was right, "There is no substitute for theory".  Nonetheless, some
folks find that practical experiments help them understand by seeing, doing
and demonstration.  I suppose that's why the Red Beads and Funnel
experiments (amongst others) were created.

Have Fun!

John McConnell
wysowl@msn.com.au