Sigma calculations

[Jean-Marie Gogue <jmegogue@wanadoo.fr>]

While reviewing some DEN messages received during the last three months, in
order to report in the quarterly newsletter of the French Deming
Association, I read the following statement:


From: Steven_S_Prevette@rl.gov  Date: Mon, 4 Jun 2001                       
Subject: RE: The rationale for substantive knowledge of underlying process  

>I have not run into field personnel that can deal with the issue of xbar and
>Range very well.  Calculating the standard deviation with the Range was
>very necessary prior to the advent of electronic calculators and personal
>computers.  Now it is not so necessary.  Of course, I have run afoul of
>the discussion of range based calculations versus statistical (sigma)
> calculations.  Even Dr. Shewhart acknowledged that the statistical sigma
>calculation was the most "powerful", but it was impractical in the days of
>mechanical adding machines and slide rules.

J-MG:

20 years ago I believed that Sigma(n-1) given by an electronic calculator
was as good as mRbar for the UCL and LCL calculation. I was intrigued
however by the fact that WED used the mRbar method in "Quality,
Productivity and Competitive Position" (prior to OOC). I also believed that
Shewhart had adopted the Range method instead of the Sigma(n-1) method
because the latter was impractical in his days.

I realized that I was wrong when I talked about this with Henry Neave and
Don Wheeler (in 1987 or 88).

Perhaps the DEN is not the best place for explaining this. Let me just
advise you to read Wheeler.

Best wishes,

Jean-Marie Gogue
Versailles  France

jmegogue@wanadoo.fr