What is SPC? Isn’t this for industrial processes?
No, Donald Wheeler makes emphasis in Understanding Variation: the book is literally for administrative and managerial data too. And in Making sense of data. SPC for the service sector.
How big should the sample size for the averages?
On Making sense of data. SPC for the service sector.
When should we recompute limits?
It’s the concept of “runs” (near the limits or above the average)
What are the statistical guarantees using this?
What about stationarity, trends?
Why not just using mean and std?
What alternatives do we have?
Anything in anomaly detection
Sonic says “say no to drugs”.
How the limits are computed
Pragmatic, non-precise. Biases
My limits are very wide!
What’s the right level of aggregation?
“These limits are far too wide to be useful.” Are they? It is not the limits that are at fault but rather the noise in the data. The limits in Figure 10.7 are based upon the data given. These limits are wide simply because these data are full of noise. The process behavior chart limits are the Voice of the Process. Whether or not you like the Voice of the Process does not matter to the process