Showing p-value for parameter estimates in choice model report

[#1]

Hi,

I'm using the choice model platform. I notice there are "estimate" and "standard error" in the "parameter estimate" report, but I'd also like to know the p-values for my parameter estimates. How can I do that?

I see one can double click on the cell to set column numeric format and do conditional highlighting by p-value, but JMP highlights the estimates and the std err in different ways (as shown in previous picture), which is very confusing to me. I want to know the exact p-values so I can report them. Thank you!

[#2]

I am not familiar with this platform, but generally, right-click on the table (e.g., Parameter Estimates), and select Columns in the menu that appears. You should see a sub-menu with the available information. Check marks confirm columns that are currently visible.

[#3]

Hi Mark,

Thanks for reply. I tried clicking on "columns" but it does not include p-value.

Any other ways I can try?

[#4]

That result means that the p-value is not available for these estimates. There is no hypothesis test available. I suggest that you contact JMP Technical Support (supuport@jmp.com) and add this to the Wish List area of the JMP Community as a new feature for the choice analysis.

[#5]

Hi @Mark_Bailey ,

Will do that!

But still, I appreciate if you know the color scheme used here in 'column numeric format'.

As I select 'p-value' in 'conditional highlighting', some of the estimates are colored in orange, some in red, and some remain black.

Does JMP color the numbers by its p-values? And what do orange, red stand for respectively?

[#6]

Select Preferences from File menu on Windows or JMP menu on macOS. Select the Reports group.

Select Manage Rules next to Show conditional formatting.

Select PValue and click Edit.

[#7]

I assume you are running a choice design experiment, but I can't tell by what you are showing us.  You might want to look at the way probabilities are calculated for these type of experiments:

https://www.jmp.com/support/help/en/16.2/?os=mac&source=application&utm_source=helpmenu&utm_medium=a...

I also can't see the data set or the associated model. It is possible you have no DF's to estimate the error.

[#8]

Hi @statman ,

Thanks for reply!

I'm running a choice model on the choice model platform. This choice model is run by JMP automatically on the platform and I think it's a MNL model. The consists of main effects and interaction terms as listed in this picture. It also shows DF's for each term in the model.

There are 64 rows on the parameter estimate report, so I have 64 parameters to be estimated. My data set has 12,960 rows (720 respondents * 6 choice sets * 3 profiles), which should be large enough.