Rethinking and Visualizing What a “City” is

Project description: As I sat in a presentation about Jacksonville, FL, where they touted their rank at the top of the largest (geographic) American cities, I thought about how we can define cities when political boundaries lead us astray. To me, using an objective density metric could give people a better idea of where “real cities” are in the United States. To accomplish this, I wrangled Census data, created an interactive dashboard, and published it online.

You can access it here: https://mtdunst-how-city-is-your-city.hf.space

Behind-the-scenes

1. Data wrangling

While the tidycensus package for R is incredible useful, it lacked the ability to easily query at the tract level with attributes for each tract’s metropolitan area. To navigate around this obstacle, I created a nested set of loops that ran through the top 40 largest MSAs and got the needed information at the tract level.

2. Creating a dashboard

While I initially considered making a static image of a city, I decided that people would be much more interested in being able to look at their hometown and see which parts of it were a “city” or not; they could relate their own experiences to what this mechanism said. Using Panel and Altair in Python, I created a dashboard that lets you choose a city and compare it to bar charts of various metrics.

3. Publishing a website

Using interactive functionality is quite simple within a Jupyter Notebook, but getting it out into the world is much trickier. Using the Docker framework for Hugging Face, I created my own site that hosts this dashboard. To give a sneak peek, I made a quick screen-recorded video showing the dashboard’s functionality in a LinkedIn post.