Look closely at data about neighborhoods, towns, or small geographic areas. Ask a question about the data, and create a set of visualizations responding to that question. (Work with non-spatial visualizations like line graphs, bar graphs, and proportional area charts.)
Start by finding data sets about that talk about places. You might look at demographics of neighborhoods within DC, a season’s box scores of high school baseball teams in your home county, or graduation rates for womens colleges around the Northeast.
This project is the first of two, related assignments. In this project, Neighborhoods, you’ll be making non-spatial visualizations like line graphs, bar graphs, and proportional area charts. In the next project, Spatial, you’ll look more closely at the location-based facets of your topic, creating a series of maps to accompany your work from Neighborhoods.
A few examples of pieces that mix non-spatial visualization types with maps (mostly thematic/data maps, in these examples).
- Mix of visualization and mapping in the FiveThirtyEight What Went Wrong in Flint
- Bar graphs, locator maps, and dot density maps in Amazon Doesn’t Consider the Race of Its Customers. Should It? at Bloomberg
- Slope graphs and choropleth maps in Away From Cities, Into Suburbs
- See also data placed on a simplified map of London in London Squared by After the Flood
1. Explore datasets and ask a question
Choose your own datasets. You might return to one of the datasets we used in previous assignments, find something yourself, or try a new one. A few possible data sources follow.
Census and related
- Census data, particularly from the decennial censuses and the annual ACS. A good place to start exploring is Census Reporter, which provides ready access to recent data.
- The Urban Institute’s Urban–Greater DC Data Explorer gives ready access to Census/ACS data for the District, Maryland, Virginia, and West Virginia, and to DC government sources. You can choose geographies, including counties, census tracts, and DC advisory neighborhood commissions.
- The official Census Bureau data site underwent a redesign last year and is more difficult to use.
More specialized datasets
- The Robert Wood Johnson Foundation’s County Health Rankings & Roadmaps site provides information about health, healthcare, and environmental health factors for counties across the country.
- The American Time Use Survey from the U.S. Department of Labor, Bureau of Labor Statistics, offers an astonishing, granular look at how Americans spend their time.
- Want to find out a university’s budget, graduation rates, or staff salaries? Or more? Look at the Department of Education’s IPEDS data.
Other ways to find ideas
Perhaps, go read the paper; dig up the original studies if need be.
Help with data wrangling
I’m happy to sit with you to help you find datasets that will help you answer your question, so far as I can. (For non-spatial data, I’m best equipped to talk about the ACS and current Census data. I also might point you toward a subject matter librarian.) I may also be able to convert data that are in formats that are not amenable to analysis, either because of organization or file format. Email me and schedule an appointment, or come by office hours.
2. Sit with the data
Start exploring the data, producing rough visualizations to help you understand what’s going on, and arrive at a question to explore. Then examine that question in more depth.
3. Design studies
Start experimenting with ways of visually presenting your data. Think of this as sketching—try different graphic forms, scales, and methods of encoding. At some stage, you’ll also need to consider how to fit this onto the page.
For this project, you need to sketch out several different approaches, with a small but representative amount of real data, but you do not need to develop multiple versions to the final stage.
4. Final design
Polish your rough design. You might want to produce new base graphics from a spreadsheet/stats program. Move on to a more visually expressive and typographically-oriented environment to prepare the final graphics—perhaps Illustrator, another drawing program, or (yes) working by hand.
- Trim size: open
- Color or grayscale
- The reader is generally curious person with at least a high school education, holding your piece at a normal reading distance for a book or a magazine.
- Encouraged: present some time-series data.
- Improve your graphics beyond the output from a spreadsheet or a visualization tool.
- Include your name, a title, and a concise introduction.
- Provide labels for key data points.
Include notes about your sources, enough that we know where you got the data from. “Data from 2012–2017 American Community Survey estimates” would be a good answer. For your own purposes, save the exact web addresses so you can go back.
Tools and methods
Use what you like. For manipulating or analyzing data, and for preparing rough visualizations, a spreadsheet or Workbench would serve well. You might want to create your base visualizations using RawGraphs or Flourish—but edit and improve them in Illustrator, another drawing program, or by hand.
For background, return to chapters 4 and 5 from The Truthful Art.
Typography, and font suggestions
This project leans on good typography as much as on the graphic side of information graphics. If you want type tips, look to the “Letter” and “Text” sections of Ellen Lupton’s Thinking with Type site, or check out the book.
Choose legible, well-drawn fonts.
Free/open source suggestions: Source Sans Pro, IBM Plex, Barlow, Cormorant Garamond, Libre Caslon, and Libre Baskerville. (More libre fonts.)
Commercial suggestions: Adobe Minion, Adobe Garamond, Caslon, Frutiger, Trade Gothic, Franklin Gothic, Myriad, Meta, DIN, Helvetica, Jenson, Archer, Gotham, and Whitney.
- “Dataset” doesn’t mean “gigantic dataset”! There’s no minimum or maximum, just start and end dates for the project. You can choose a tiny or simple dataset and invest your time in exploring design.
- You will almost certainly need to show several visualizations on one page. A single, monster visualization will sink under its own weight. A lone, featherweight visualization will underwhelm. Give us substance.
- Truly, the labeling defaults on stats/spreadsheet software are terrible. Do not label everything; we won’t be able to see all the text and no one will care. But label the values that matter.