Data visualization

SDS 236

Benjamin S. Baumer

Smith College

Mar 31, 2025

Data visualization

How Working-Class Life Is Killing Americans, in Charts

https://www.nytimes.com/interactive/2020/03/06/opinion/working-class-death-rate.html

• What do you notice?
• What do you wonder?

How Does Your State Make Electricity?

https://www.nytimes.com/interactive/2018/12/24/climate/how-electricity-generation-changed-in-your-state.html

• What do you notice?
• What do you wonder?

What makes a good graph?

Some answers:

• Tells your story
• Reveals the data
• Illuminates, not confuses
• Grabs interest
• LOW ink density (ie “not real complicated”) …
  • … for HIGH data density

What makes a great graph?

• Viewer sees themself in the graph
• Surprise

Good viz

Sorry, no rainbows

• ❌ 🌈 NO RAINBOWS!!

• Why rainbow colors aren’t the best option for data visualizations

• How The Rainbow Color Map Misleads

Elements of data graphics

Elements of data graphics

• Visual cues
  • position, length, area, etc.
• Coordinate system
  • how are the data points organized?
• Scale
  • how does distance translate into meaning?
• Context
  • in relation to what?
• Faceting / small multiples

Perceptual heirarchy

Do’s and don’ts

Do:

• Mind the perceptual heirarchy
• Put contextual info on the plot
• Make an appropriate color choice
• Match data dimensions to visual dimensions

Don’t:

• Don’t use color when length will suffice
• Don’t use pie charts!
• Don’t use redundant mappings unless you have a reason
• Don’t map area to a 1D quantity

Read more about this!

Two more books

]

Tufte’s mantra

Color

Three color schemes

• Sequential

  • Single hue
  • Multi-hue

• Diverging

• Categorical

Big Idea

Color choices matter and there are better and worse choices

Sequential, single hue (continuous)

• Map saturation to numeric variable

• People interpret high saturation as more

Sequential, single hue (discrete)

• Map saturation to numeric variable

• People interpret high saturation as more

• No more than 5–6 levels

Sequential, double hue

• Map hue to numeric variable

• Generally, no more than 2 anchor colors

• People interpret darker hue as more

Diverging

• Two colors used to indicate extremes of a range

• Neutral color in the middle

• distinguish between zero and missing

Categorical

• Different color = different category

• choose colors that are

  • perceptually distant
  • roughly the same saturation

Built-in color palettes

• Cynthia Brewer
  • RColorBrewer package
  • scale_color_brewer() function
• viridis
  • Comparison

Workshop

bbplot from the BBC

Improving Economist graphics (1)

Improving Economist graphics (2)

Improving Economist graphics (3)

Improving Economist graphics (4)

Improving Economist graphics (5)