Subscribe Bookmark RSS Feed
Byron_JMP

Staff

Joined:

Apr 26, 2012

Interactive HTML Nominal Logistic Regression Profiler Figure

The Nominal Logistic Regression Report has a  Profiler for showing the probability of a binomial outcome as a function of your model effects. This is a great figure, but when its exported using the "Interactive HTML" option, the figure is static. These are some quick notes on how to get around that problem.

Screen Shot 2016-12-15 at 4.07.57 PM.png

 One of the things I love about the nominal logistic Profiler is that it shows the probability of one level over the other in a slick little boxed Y axis scale. Unfortunately this is key to why the figure doesn't export. So if you're willing to accept a Y-axis that shows the probability on a 0 to 1 scale for one of the two levels, which isn't a bad compromise, then give this a try next time you want to publish that interactive logistic regression figure.

 

 

Method 1: Remove the Response Probability Property before Graph>Profile

Step 1: In the Nominal logistic regression report, use the red triangle option to save the probability formula to a column formula.

When you do this, 4 columns will be saved to the table.  We're interested in one of the two that say something like Prob[0].  

Step 2: This column has an asterisk next to it, click the asterisk and the column properties dialog box will pop up.

Step 3: Remove the "Response Probability" property. No more nifty y-asix box for probabilities.

Step 4: Go the Graph>Profiler, select the column you just modified (e.g. Prob[0]) for the formula. Click the Expand intermediate columns box (super important)

Step 5: Save your awesome new logistic regression figure as interactive HTML (MAC: Export Interactive HTML) and you're good to go.

note: this is for JMP 13, older versions might be different

 

 

Method 2: Save Expanded Formulas from the Profiler, Profile Expanded Formula

Step 1: In the Nominal logistic regression report, use the red triangle option to save the probability formula to a column formula.

When you do this, 4 columns will be saved to the table.  We're interested in one of the two that say something like Prob[0].  

Step 2: Use Graph>Profiler on one of the Prob columns. Click the Expand Intermediate Columns box (super important)  before clicking OK

Step 3: In the Profiler window, use the red triangle menu to Save Expanded Formulas.

Step 4: Use Graph>Profiler on the Expanded Formula column.

Step 5: Save your awesome new logistic regression figure as interactive HTML (MAC: Export Interactive HTML) and you're good to go.

 (thanks to Bill Worley for this tip)

Article Tags