How to create choropleth maps with percentile-based colorbars

📊 Plotly Python
1

Hi everyone,

As part of my Python for Nonprofits project (which remains a work in progress), I’ve written a notebook that shows how to create choropleth maps with percentile-based bins. This approach helps prevent outliers from skewing your map’s color scale. Although the notebook is still a work in progress, I thought I would share what I’ve written so far in case it would help others who are trying to create percentile-based maps.

Here’s the original map, with a linear colorscale, that my code shows how to update: (Note that a few outlier counties with particularly large population changes have caused the other counties to appear relatively homogeneous.)

sample_linear_map
sample_linear_map1920×1080 189 KB

And here’s the updated map. Note that the colorscale has been updated to show percentile-based increments (~0th percentile, ~10th percentile, ~20th percentile, and so on) rather than linear increments. As a result, the map is much more colorful and easier to interpret.

county_pop_pct_growth_2011-2021
county_pop_pct_growth_2011-20211920×1080 218 KB

In order to make these updates, I first changed the color argument of my map to a column that stores the percentiles for each value rather than the actual values. (Such a column can be created via Series.rank(pct=True).) I then modified two values of the map’s colorbar:

  1. colorbar_tickvals, which specifies where along the colorbar (e.g. from the lowest percentile to the highest) text labels should be added. (I set this property to specific percentile ranks within my dataset.)
  2. colorbar_ticktext, which specifies what text should be placed at those values. (I set this property to the actual values within my dataset that corresponded to the percentile ranks passed to colorbar_tickvals.)

The code is released under the MIT license, so feel free to repurpose it for your own projects (including commercial ones).

As time allows in the next week or so, I’ll consolidate this code into a function so that it can be more easily applied to different projects. I’ll plan to add that function to the same Mapping section of Python for Nonprofits that contains the notebook to which I linked above.

2 Likes
2

Hi!! @KBurchfiel. You don’t need me to tell you that it is a fantastic work, but I didn’t want let pass the opportunity to thank you for your collaboration. I guess that everyone of us know how time-consuming is working on this kind of project and same as you, there are ton of great contributors in this community (:star_struck:)… anyway. I 've just bookmarked and I’m waiting for your final release!
Thanks again…

2 Likes
3

I love this. Such clear and helpful code. I look forward to seeing how progress with this project, @KBurchfiel .

The second choropleth map is indeed much clearer.

1 Like
4

Thank you, everyone!

To make the code a bit more portable, I’ve added a mapping function into this Python file. This function has also been updated to accommodate data labels (which are useful for US state-level maps) and to generate percentile data automatically.

The Jupyter notebook that I shared in my original post now shows usage examples for this function along with some additional maps. I plan to add some additional documentation to the notebook in the coming weeks.

1 Like