nav[aria-label="Primary Navigation"] { padding: 0; & ul { list-style: none; width: 100%; display: flex; flex-direction: row; justify-content: start; align-items: start; gap: 30px; padding: 0; & li { margin: 0; } & ul li { list-style: none; } } }

Confidence Intervals for a Binomial Proportion in a Graph?

Peter_Bartell

I've got a customer that wants to create a chart analogous to a Fit Y by X continuous variable report with confidence intervals on the proportion similar to a Box and Whisker type chart where there is a plotting symbol at the estimated proportion, and the whiskers extend vertically from the point estimate to the CI boundaries for the proportion. The x axis are just collections of samples over some sampling frame. The ideal is a solution a 'no scripting required' solution. Can either Fit Y by X or maybe Graph Builder be cajoled into something like this? I can see a mulit step solution...estimating the parameters and CI's in the distribution platform...then putting these into another data table to drive Fit Y by X or Graph Builder...just wondering if I'm missing a workflow that could run from the original data table in one fell swoop?

Accepted answers

All comments

yusuke_ono

I am a tester, and I don't know the real world.So the following answer might be inappropriate.

I think you are correct.
As far as I know, there is no option to draw a confidence interval for binary data in each group.
I think that the only way is ... to make a script .

The follwoing is a Proof-Of-Concept (POC) script. This is a very primitive and naive script, but I hope this script will show you how easy it is relatively.

https://community.jmp.com/servlet/JiveServlet/downloadBody/8337-102-1-9489/binomial%20graph.jsl

The script is a little bit long, but the main part has only about 10 lines. The script invokes the Distribution platform, outputs the results to a data table, and draw the graph by the Graph Builder platform (as you said).

Run this script, select columns, and click one of response levels displayed in the Local Data Filter.

Note that this script can work only on English JMP. I made this script on JMP12.2 (I have not checked on the other versions).

This is not an appropriate answer for your question, but I think the only way is to make a script ... as usual.

Should I enter this request for the Categorical or the Contingency platform into the Defects as a request from user(s)?

jules

Hi peter.bartell@jmp.com

What does this customer mean by "chart analogous to a Fit Y by X continuous variable report?" Do you mean the Distribution report for a continuous variable? If " The x axis are just collections of samples over some sampling frame" means that the X axis is categorical, that sounds like something Distribution can handle. Distribution has the option to put standard error bars on the observed frequencies under the histogram options under the red triangle. Is the output below like what your customer wants? If I'm missing the question an example graph or even a sketch would help.

julian

10103_Screen Shot 2015-10-01 at 6.54.52 PM.png

jules

Hi again, peter

Never mind, I think I know what your customer is asking for (but I don't think I can do it in JMP yet so I'm just mocking this up for now). What about the graph below?

10104_Screen Shot 2015-10-01 at 7.05.31 PM.png

jules

Hipeter.bartell@jmp.com

The graph below didn't require any scripting, but did require some creation of combined data tables and some stacking.

10105_Screen Shot 2015-10-01 at 7.32.25 PM.png

Would that do? If so I'm happy to write out steps.

julian

Peter_Bartell

Julian: Thanks for your contributions on this thread. I too was able to cobble together something similar to what you did too. Ultimately, what the customer was really asking about was was there some sort of hidden drop down, buried in some platform report, somewhere, which could do this sort of thing. They had a home grown solution too...just wondering to me out loud about the issue, and I didn't have an off the shelf answer so I thought I'd post the question on the SE space.

jules

After all my posting I went back and read your original message again and realized that you had already found a solution but were looking for something straightforward! Sorry about that!

Peter_Bartell

yusuke: I don't think you need to enter this issue as a defect. It's not a 'defect' in that sense of the word. A new feature maybe...but it's just one customer asking for this from me so unless there was a groundswell of interest from other customers I can't see spending the effort on this issue. But that's just my opinion.

yusuke_ono

I also read back your original question and my answer, and I realize that I ddin't answer anything for your question.

I am sorry for my inappropriate answer.

I cannot find a better way than programming with JSL or your method...