CartoDB Editor: Bubble map > Classed option

[makella]

Cartographically speaking, there are two types of 'Bubble' maps proportional symbol (unclassed) maps and range graded (classed) maps. This issue is in reference to range graded, I will add another, separate issue, related to proportional symbol.
On range graded maps, each symbol size is determined by the class it falls into. This is similar to the choropleth points but uses size instead of color as the visual variable.
Right now in the Editor, there is the option to choose a classification method, but there isn't an option to choose how many class breaks the data should have. By default, the Editor is creating 10 class breaks and the result is little to no visual variation in the graduated symbol map because the size difference between each symbol within 10 classes is really small. In order to be a true range graded map, we need to add the option to choose the number of classes. We also need to modify the legend design for this type of map and the CartoCSS.
The proposed workflow, modified CartoCSS and legend design are all described in more detail below. This would be our Bubble > Classed option.

Proposed Workflow:

• Select Column to display
• Choose Quantification method
• NEW: Choose number of Buckets (default to 5)
  • I think having 3,5,7 like the choropleth option is good so the user can select the number of classes that best help show the variation in their data.
• Choose min/max symbol size
  • I think we should change this to be something that will produce non-integer numbers for the default of 5 classes because it seems that having non-integer provides more distinction between symbol sizes. This needs more testing so for now, I'd say we can stick with the defaults.
• NEW: Order data: ascending or descending
  • This is an important step, giving users the option to have the smaller data values draw on top of the larger ones so they aren't hidden
• Bubble fill color
• Bubble stroke
• Composite operation
  • I'd like to spend a little more time looking at these because I think there are ones that help see overlapping data points and others that start to introduce different colors into the map that can't necessarily be explained in the legend

    Enhanced CartoCSS view

• In the CartoCSS view (true for all classed data's CartoCSS), we should have the classes the data fall into clearly written in the CartoCSS. This will allow people, who wish to, be able to more easily define 'Manual' classification steps for their data.
  A simple example:

#dataset {
  marker-fill-opacity: 0.8;
  marker-line-color: #FFF;
  marker-line-width: 1;
  marker-line-opacity: 1;
  marker-fill: #FFFFB2;
  marker-allow-overlap: true;

  [ estimate_total >= 251][estimate_total<=800] {
   marker-width: 25;
  }
  [ estimate_total >=51][estimate_total<=250] {
   marker-width: 20;
  }
  [ estimate_total >=26][estimate_total<=50] {
   marker-width: 15;
  }
  [ estimate_total >=8][estimate_total<=25] {
   marker-width: 10;
  }
  [ estimate_total >=1][estimate_total<=7] {
   marker-width: 5;
  }
}

Enhanced legend

For a classed bubble map, the legend should show class breaks with associated symbol size. Important here is that if the values of the data aren't as clean as they are in my simple example above, that the numbers should be rounded for the purpose of the legend.
Example legend:

cc @andrewxhill

[javisantana]

We are working on the new embeds so @josecruz might should take a look at this to see if the bubble design fits the requirements.

A technical question, where is better to use classed/unclassed bubble maps. I'm thinking on pecan

[makella]

@javisantana good question and something that I was thinking about too.

In my opinion, this is related to a few different things:

• the distribution of values in the data,
• the unit at which the data were collected,
• the viewing scale of the map and,
• the theme of information being mapped.

From the pecan angle, I think there are some patterns we can look for in the data to guide users to make better decisions -- the biggest thing here is to make sure we help people make readable maps and not just maps with a bunch of bubbles!

Since there is more variation in symbol size on unclassed maps, we don't want to use this option if there is little variation in the data because the symbols on the map would look homogenous. On the other hand, if there are greater variations in the data, the symbol size would reflect that and make for a more interesting map. BUT there are instances where I can think of using an unclassed map with a smaller range of values at larger map scales.

If there isn't too much variation in the data, then a classed map would likely be the best option because that would help the map reader best understand the patterns in the data. Also, with classed maps, there are fewer symbol sizes to interpret which again depending on theme and scale could make the map a lot easier to understand.

Let me research this topic a little more and find some examples to better illustrate the points above.

[makella]

How are we currently (and in the future) determining the size difference between symbols for classed bubble maps? I found some resources that can help and that I can test if needed.

[Yeye]

Makella can you share the code for the Example Legend you showed above?

[alonsogarciapablo]

There has been no activity on this issue for more several months. We are closing it. If you think this still needs to be addressed please open a new issue.