(attach to prior fragment -- 1.5 sigma shift happens)
Reviewer: David Hsllowell dhall@hallsystems.com from Portsmouth, NH 5/9/2000
(..attach to prior segment..inadvertently a keystroke sent it off)

When a subgroup plots with a 'shift' of 1.5 sigma we don't really know that the process has shifted -- the sampling distribution will produce these 1.5 sigma 'shifts' without any change in the underlying process. That, to me, points to the great injustice that snuck into the six sigma history -- somehow these shifts got interpreted as being mysterious properties of characteristic of 'typical processes'. A bowl of beads will generate these '1.5 sigma shifts' with subroups around 4 or 5. That was the explanation that I got back from the people in Motorola back in 1990 - before the mystery of the shift had been propogated. So, my humble take on all the above is -- under control chart management, apparent shifts of more than 1.5 sigma will trigger special action - so discussion of defect rates is moot. Otherwise, we could assume that the worst case defect rate would attach to the 'just before shutdown' condition at about 1.5 sigma. That alone simplifies all the conversion charts and soul searching about 'do *we* have the 1.5 sigma shift' like some ghost in the machine. Jon did such a good job disarming a lot of that, in my opinion, false and harmful bull-shift that I had try to contribute this supplementary stake through the heart. I hope it helps.