ANOM JMP v's Wheeler

[#1]

Hi I am trying to replicate an ANOM that Wheeler did in Understanding Industrial Experimentation p60 (details attached)

I have attached the data set, I get different control limits for some reason, I set the alpha level at 0.10

Can you advise

[#2 · Accepted solution]

The JMP help notes that JMP implements the methods from Nelson, et. al (2003).  link to JMP ANOM help 

The methodology in the Nelson book differs slightly from Wheeler's, for instance the measure of error differs. Therefore, the ANOM decision limits calculated in JMP will differ from those in the Wheeler book. You can find the Nelson ANOM book here (ANOM Book ).

For the equal case for means the decision limits are:

[#3]

Thanks Karen, I asppreciate your response here

[#4]

Hi Karen, I just looked at the book, your one of the the authors. So which is the best method and why ;0) 

[#5 · Accepted solution]

Wheeler uses ranges, JMP/Nelson uses standard deviations. Generally they are going to give you "similar" results. The key is plot your data, understand your measurement systems, use ANOM decision lines to guide your decision making. Since you are using JMP use the ANOM charts in JMP.

[#6]

Thanks, that really helps understanding the difference.  I have dealt with SD and ranges with the control charts.  The Shewhart chart and the Levey–Jennings charts for calculating control limits, where the Levey–Jennings uses SD.  Its important to understand these differences.

[#7]

Another important distinction between Shewhart charts and Levey-Jennings charts is that the former types rely on short-term variation that is internally estimated in Phase I. The latter type often uses externally determined long-term variation. A clinical laboratory scenario, from which they originated, is given the SD or control limit information by the manufacturer of the assay, which they base on a precision study that intentionally includes long-term variation.

[#8]

Hi Mark

I must admit I do struggle with  Levey-Jennings charts.  When the process chart is stable the control limits equate to Shewhart chart limits, but when the process is not in control (as mesured by the Shewhart chart) the control limits are more forgiving (wider).  Now I undertand that the Westguard Rules should then kick in, but very often people do not use these Rules, so the sensitivity of the Levey-Jennings chart is diminnished.  A Strong signal becomes a Weak signal.  The standard deviation component of the Levey-Jennings chart assumes the data is consistent, stable and attempts to fit these wider control limits to all the data.