Subscribe Bookmark RSS Feed

Bivariate Histogram

joe_schaar_gmai

Occasional Contributor

Joined:

Aug 9, 2016

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
ron_horne

Super User

Joined:

Jun 23, 2011

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

joe_schaar_gmai

Occasional Contributor

Joined:

Aug 9, 2016

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

Joined:

Jun 22, 2012

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

Joined:

Jul 22, 2014

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! 

joe_schaar_gmai

Occasional Contributor

Joined:

Aug 9, 2016

Phil,

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

-Joe