Shift Ratio or Percentage Change

[#1]

Hi Everyone, 

I am looking for a function on JMP software that is called shift ratio or percentage change (it is not available with either of these names under statistical index). It is normally used to detect how much the outlier has shifted in case Nelson 2 rule is not met in a control chart. 

Could you please kindly advice if you used this function on JMP, and if yes, how would I find it?

Thank you in advance 

@Ben_BarrIngh 

[#2]

I know that Process Screening platform has Shift Detection but I'm not sure if that is what you are looking for Process Screening Platform Options (jmp.com)

MSA platform also seems to have Shift Detection Profiler (jmp.com)

and then there is Example of Operating Characteristic Curves (jmp.com). 

I found these by searching from JMP Help https://www.jmp.com/support/help/en/17.2 and from scripting index using Shift as search term. There is also Search JMP which can be used to search for things in JMP (Help / Search JMP)

[#3]

Thank you very much for your detailed response. I have checked the links and descriptions you have provided. Unfortunately, that's not what I am looking for. 

It also goes by the name shift factor which is calculated to quantify the magnitude of the shift in case of Nelson 2 violation. 

[#4]

Hi @ZA : I've never heard of this, by any of those names anyway.  Sounds like maybe it calculates ratio of sample means: i.e.,

Mean1[n=9 for the nine in a row above/below overall mean]/Mean2[overall or rest of data], or some such calculation? Does something like that ring a bell? 

This assumes by Nelson Rule 2 you mean "Nine points in a row are on the same side of the mean"

https://en.wikipedia.org/wiki/Nelson_rules

[#5]

Hi @MRB3855 ,

Thank you very much for your response, yes the calculation is something like what you have written and as you mention it relies on Nelson 2 rule. As I understand this isn't automatically calculated by JMP. 

[#6]

Hi @ZA : Yeah, it may not be any sort of "standard" calculation; that said, it could be informative as it could be seen as an estimate of how much of a mean shift may have occurred. You may have to calculate it yourself; via some slick JSL (I'll leave that to the experts), or just create another column in your data that identifies the rogue n=9, then calculate the summary stats by that new column? Or, you could use the Y by X platform and choose that new column as X (so long as it is a character/nominal factor).

I'm sure there are other ways...    

[#7]

What is the source of the data you want to analyze?  Is it a time series? (or some other rational series?). Are you looking to plot the data and apply rule 2?

perhaps this will help?

https://www.jmp.com/support/help/en/17.2/?os=mac&source=application#page/jmp/options-panel-and-right...