Re: Rule of thumb for SPC

[Rip Stauffer <ripstaur@vabch.com>]

Ian,

There are any number of things that could be said. I have some
questions:

When you ask, in question 2, about the "run chart in control," are you
talking about the R chart in an XbarR chart? If you are, the scenario
you describe could be caused by subgrouping unlike things in the x
chart. I recommend you run a separate chart on each of those things. You
will find a useful description and a great discussion of this in Wheeler
and Chambers, starting on (I think) p. 104--"the Ball Socket Data."

In question 1, why are 20 percent of the points in the R chart outside
the UCL? You could exclude them all, but until you know what is causing
them, you will be unlikely to improve your process, and any limits you
construct will not necessarily reflect the underlying process. Another
good thing is that once you know what is causing them you can remove the
cause and that will remove them anyway. THEN recalculate your limits.

If you want to send your data, I'd enjoy taking a look at it. Sounds
like an interesting data set.

Rip Stauffer
Naval Leader Training Unit
ripstaur@vabch.com
===================================================================