RE: British Election turnout - one of the lowest: Special case or

["DeRoeck, Rich" <RDeRoeck@alphaind.com>]

  >>Interestingly, Wheeler's constants (from
"Building Continual Improvement") generate narrower limits than does +/-
3stdev. I seem to recall Wheeler mentioning in the book that his reasoning
for using the constants (rather than sigma) is complex and likely difficult
to follow for someone who is not a statistician.  Any DENizens who could put
the rationale in layman's terms?<< 

Loren,

How was sigma calculated? Wheeler makes the important point that it's HOW
sigma is calculated (average or local dispersion statistic vs. a global or
overall calculation) that can effect the width of the process limits. He
writes, "Limits for measurements are always based upon the average or median
of a set of k dispersion statistics". The process limits will be inflated if
there are signals (assignable causes) in the data. For more on this subject
I recommend reading from his book "USPC" Chapter Nine, Section 9.2 XmR
Charts Done Right. 

Hope this helps,

Rich DeRoeck