cancel
Showing results for 
Show  only  | Search instead for 
Did you mean: 
Try the Materials Informatics Toolkit, which is designed to easily handle SMILES data. This and other helpful add-ins are available in the JMP® Marketplace
Choose Language Hide Translation Bar
Julianveda
Level IV

"Actual by Predicted plot" - How confidence intervals are really calculated ?

 

Hello Community,

 

My Question:

I have a question concerning how I can access the confidence intervals of the Actual by predicted graph below:

Julianveda_1-1689492079621.png

 

 

I attach the jmp file if you want to try yourselves. Now I am gonna tell you what I have tried up to know:

 

What I have tried:

I have save the columns “mean confidence limit formula as shown below:

Julianveda_2-1689492079628.png

 

However, when I try to reproduce the “Actual by Predicted plot”, the confidence intervals don’t look similar :

Actal by predicted plot:

 

Julianveda_3-1689492079629.png

 

 

Graph done in Graph builder with the saved columns (predicted, lower and upper confidence intervals):

 

Julianveda_4-1689492079632.png

 

 

So the saved columns clearly do not correspond to the confidence intervals in the “actual by predicted plot”

 

However, I realized that the saved confidence interval columns are actually the confidence intervals that appear in the profiler. Look at the two pictures below, one from the profiler and the other one  built in the Graph builder:

 

Julianveda_5-1689492079634.png

 

 

Julianveda_6-1689492079649.png

 

 

This is interesting, but my original question is still unanswered:

 

How the confidence intervals from the “actual by predicted plot” are calculated ?

 

I also tried calculating them  by myself in excel using the statistical equations for this, but still the confidence intervals I find are different from the “actual by predicted plot”:

 

 

 

Confidence interval calculated with t=0.05

Julianveda_7-1689492079650.png

 

 

Very different from

Julianveda_8-1689492079651.png

 

 

Conclusion:

Neither in JMP nor by my own calculation I have not been able to find the right way in which the confidence interval in the “actual by predicted” are actually calculated.

 

Could please someone help me to understand ?

 

Thank you,

Julian

 

 

 

 

 

3 REPLIES 3
jthi
Super User

Re: "Actual by Predicted plot" - How confidence intervals are really calculated ?

Usually good starting point is using Help tool found from JMP's toolbar (?). Hover over the part you are interested in the analysis while the ? tool is enabled

jthi_0-1689515834193.png

and it will open help page related to that https://www.jmp.com/support/help/en/17.0/#page/jmp/effect-leverage-plots.shtml?os=win&source=applica... . Use this as the starting point, then if you don't find what you are looking for check the left side of help page and look for "Statistical Details...."  https://www.jmp.com/support/help/en/17.0/#page/jmp/statistical-details-for-leverage-plots.shtml#ww15... . Hopefully these provide at least some additional information

 

-Jarmo
Julianveda
Level IV

Re: "Actual by Predicted plot" - How confidence intervals are really calculated ?

Thank you @jthi  for the links. I have already seen the one about the leverage plots but not the one of the statistical details. I'll have a look into that to see if it can help me. Thank you also for the tip of (?) tool; it's may be simple but I did not know about it

 

Victor_G
Super User

Re: "Actual by Predicted plot" - How confidence intervals are really calculated ?

Hi @Julianveda,

 

The questions about "Actual vs. Predicted" plot are quite common in this Community, you can also look at some previous posts with the provided answers, it may already solve some of your questions
Example here : https://community.jmp.com/t5/Discussions/Actual-by-predicted-plot-confidence-region/m-p/641519 

 

I hope this will also help you,

Victor GUILLER

"It is not unusual for a well-designed experiment to analyze itself" (Box, Hunter and Hunter)