Here are some questions that were emailed to me by a reader….
#1 – If I have a run of 8 signals and need to recalculate my UCL and LCL, do I start the new limits with the first of the 8, the last of the 8 or the first one after the 8?
Don Wheeler teaches that you’d generally start the new average and limits calculation with the first of the 8 data points, as that’s likely when the process / system change occurred — if you don’t have any more specific knowledge about the timing of a system change.
#2 – What dataset do I use for recalibrating? Is it a total of 25 starting with the 8? Is it just the 8?
I’d start with the first 8… and I might occasionally “refine” the average and limits with more data points as they’re added (up to 20 or 25, that’s about the same effect).
The thing you don’t want to do is to continually recalculate limits using “the last 25 data points” or something like that. You really should be starting a new calculation range when there’s been a signal or a known change to the system.
#3 – How would I interpret it if I have a rule 1 signal immediately as the first data point after setting new UCL and LCL limits?
That would tell me that the system has changed again. That’s what a signal of any type is telling us.