cancel
Showing results for 
Show  only  | Search instead for 
Did you mean: 
Submit your abstract to the call for content for Discovery Summit Americas by April 23. Selected abstracts will be presented at Discovery Summit, Oct. 21- 24.
Discovery is online this week, April 16 and 18. Join us for these exciting interactive sessions.
Choose Language Hide Translation Bar

Bivariate Histogram

Can JMP create a bivariate histogram of 2 data columns from a data set? I am thinking of something that creates graphs similar to those one can make in Matlab using the "hist3" command?

http://www.mathworks.com/help/stats/hist3.html

(see bottom figure in this link as an example of what I'm visualizing)

I don't necessarily need the fancy aspect viewpoint like shown in the figures from the Mathworks website. Having a top view with "counts" represented with color scale for the vertical dimension would be great.

Thanks

5 REPLIES 5
ron_horne
Super User (Alumni)

Re: Bivariate Histogram

hi joe.schaar@gmail.com​,

if you just need a 2D version try heat map until someone gives the more comprehensive solution.

in graph builder put the two variables on the x and y axis and click the heat map button from the top menu.

12397_pastedImage_0.png

best,

ron

Re: Bivariate Histogram

Thanks Ron. That is exactly what I was looking for!

An add-on question here...once I select a bin(s) from the heat map I see that the corresponding rows are selected within the originating data table as expected. Is there a way, separate from the obvious copy/paste approach, to a have only the selected data (both columns) populate a new data table?

-Joe

txnelson
Super User

Re: Bivariate Histogram

Concerning your addon question:

If you go to the pulldown menu

     Tables==>Subset

The platform comes up with the option defaulted to subset only the selected rows

Jim
Phil_Kay
Staff

Re: Bivariate Histogram

I think Ron's suggestion is great. You could also use the density plot option in Graph Builder if more definition would be helpful.

12401_pastedImage_2.png

You could create something like this with the 3D scatterplot:

12399_pastedImage_0.png

You need to first bin the variables and then create a table where you have frequency or N for each of the 2D bins. I did this by saving midpoints from the Distribution platform and then Table > Summary. There is probably a quicker way.

However, I can't see what this 3D plot gives you. Or the 3D plots in the link that you shared - in fact they are even worse because you can't see what is happening behind a peak. 3D plots only really work when you can spin them. Best not to add the 3rd dimension unless you have a really good reason. The colour in the above graph is probably not adding anything either! 

Re: Bivariate Histogram

Phil,

Those are great suggestions, and I like the visuals. Thank you very much!

-Joe