Re: SPC - independent data

["Dr A D Burns" <mmedia@ozemail.com.au>]

Steve,

There's a couple of points here. The first step in problem solving, is to
define the problem. A full definition of a problem may make SPC irrelevant.
It is important not to just chart for charting's sake. A control chart for
overall stock levels in Steve's warehouse example, would have been quite
appropriate and would have rang appropriate warning bells if the problem
related to detecting changes in stock levels. The system would have been
unstable, and potentially led to zero stock. Secondly, the purpose of the
control chart is to gain insight into the process. It is not necessary to
carry out other tests to determine the type of distribution or the degree of
independence of data. The control chart itself is all that is needed. Where
we already do know information about the process, we can design a control
chart to be more effective, for example by better selection of subgroups.
Finally, out of interest it's worth noting that Wheeler suggests that
additional detection rules (like 7 (or 8) in a row the same side of the
average ) increase the chance of false alarms more than increasing the
sensitivity of the chart, beyond use of the usual 3 sigma control limits.

Regards,

Dr Tony Burns
MicroMultimedia
www.q-skills.com
mmedia@ozemail.com.au