[Jacob Sagrans]

When coloring the points on a graph by a numerical attribute, the scale is problematic in that adjacent bins include the same boundary values. Check out this link, where I clicked on the middle bin in the graph legend (10 to 12 hours of sleep) and then selected “Hide unselected cases.” I expected to see all the points in the filtered graph being the same color, but they are not, as the 10 hour values are classified in the bin below the one I filtered on (4.6 to 10 hours) and the 12 hour values are classified in the bin above the one I filtered on (12 to 17 hours). This also raises a question of why we see these points, if according to CODAP, they are not in the bin I filtered in (although these points were highlighted when I initially clicked on the 10 to 12 hour bin/bar in the legend, so it seems like CODAP is maybe treating them as belong in two bins, or maybe in one bin in one context and another bin in another context?).

[Jacob Sagrans]

Oddly, if I open the link I shared now, the scale is redone and is only 10 to 12 hours, although there does still seem to be an issue of boundary values belonging to more than one bin.

[Jacob Sagrans]

This is what it looked like initially after I did the filtering. All the points shown are in the 10 to 12 bin (and are all highlighted if I click on that bin in the bar in the legend). But, as shown below, if I click on the 12 to 17 hour bin, points for 12 hours only are highlighted too.

 

Attachments:

1. 

   CODAP-binning-numeric-attribute-with-colored-points-on-graph-ex.png
   

[Bill Finzer]

Hello Jacob,

You are correct that with choropleth legends some values get lumped in more than one bucket. This becomes dramatic if you happen to be using an attribute whose values are all identical. CODAP’s implementation puts them all in all the buckets!

It would be good if CODAP treated numeric legends more flexibly, and especially good if it gave users more control over the colors.

[Jacob Sagrans]

Thanks for the reply Bill. Maybe this could be changed/improved for CODAP at some point in the future? I should see if there is something about this in the CODAP Software Pivotal Tracker, and if not, add it.

[Jacob Sagrans]

Turns out there is a lot pertaining to the color scale already in Pivotal Tracker:

“Quintile selection doesn’t adjust after hiding selected/unselected”: https://www.pivotaltracker.com/story/show/183831874

“Allow choropleth legend to reflect values in entire dataset even when only a subset is being viewed”: https://www.pivotaltracker.com/story/show/182702665

“Two categories can wrongly get assigned to the same legend color”: https://www.pivotaltracker.com/story/show/158934514

And more here: https://www.pivotaltracker.com/n/projects/1055240/search?q=color%20legend

[Author]

Posts