Re: Relation between 3 STD and approximated STD from XmR

[Alson C H Look <alook@pacbell.net>]

There has been a great deal of controversy in the tech literature concerning
the XmR charts.  The moving range is often a better estimator than the
standard deviation since the moving range has the nice property of being
resistant to "trends."  The standard deviation estimator often tends to be
inflated if the data contains any "trends" (cycles?).

When one is establishing process control it is often a good idea to use the
moving range to estimate control limits rather than the standard deviation
for the reason mentioned above.  I am assuming that rational subgrouping
(n>1) here does not make sense.  (For example, you have a chemical process
and would like to understand it.  Using a rational sample of n = 1 may make
sense.)  You still must think about your process to determine whether XmR or
Xbar and R charts make sense for the process.

Alson
alook@pacbell.net