Data Viz Considerations





Grayson White

Math 241
Week 1 | Spring 2026

Week 1 Goals

Day 1 Lecture

  • Course overview

Day 2 Lecture

  • Develop language to talk about the components of a graphic

  • Discuss considerations for good graphical design

  • Graphics consideration activity

  • PSET 0 assigned

First: P-set 0

Let’s start with the language we will use to describe the components of a graph:

The Grammar of Graphics

Background


Leland Wilkinson wrote a book called “The Grammar of Graphics”




As part of his PhD in Statistics at Iowa State, Hadley Wickham wrote the R package ggplot2, which we will use to create static graphs.

The Grammar of Graphics

  • data: dataset that contains the data

  • geom: geometric shape that the data are mapped to

    • point, line, bar, text, …

  • aesthetic: visual properties of the geom

    • x position, y position, color, fill, shape

  • coord: coordinate system

    • Cartesian, polar, geographic

  • scale: controls how data are mapped to the visual values of the aesthetic

    • EX: particular colors, linear

  • guide: legend to help user convert visual display back to the data

Geoms versus Names

  • What are the names of these graphs?

  • Focus on the shapes (i.e. geoms) and how the variables are mapped to those shapes.

Data Viz Example

Let’s practice deconstructing this graph using the grammar of graphics.

  • Geom(s)?
  • Aesthetics of the geom?
    • Mapping of variables?
  • Coord?
  • Scales?

Data Viz Example

Let’s practice deconstructing this graph using the grammar of graphics.

  • Geom(s)?
  • Aesthetics of the geom?
    • Mapping of variables?
  • Coord?
  • Scales?

Choices

For most data, there won’t be just one way to graph it. Decisions to be made:

  • What geom to use
    • Point, line, bar, …
  • For a given geom, how to map variables to its aesthetics
    • Size, location, color, …
  • For each aesthetic, what scale to use
    • Linear, diverging colors, …
  • For the graph, what coordinate system to use
    • Cartesian, polar

Let’s discuss some considerations that can help guide these decisions. But…

“Data visualization is part art and part science. The challenge is to get the art right without getting the science wrong and vice versa.” – Claus Wilke

Recommendation: Try out different options and make sure to iterate!

Consideration: Consider variable type when picking the aesthetic mapping.

What aesthetic options do I have at my disposal?

Aesthetics: Position/location

Aesthetics: Length

Aesthetics: Area

Aesthetics: Angle

Aesthetics: Shapes

Aesthetics: Color Shade

Aesthetics: Color Hue

Which represents the larger value?

From Wickham (2012)

Which represents the larger value?

From Wickham (2012)

Which represents the larger value?

From Wickham (2012)

Which represents the larger value?

From Wickham (2012)

Which represents the larger value?

From Wickham (2012)

Which represents the larger value?

From Wickham (2012)

Consideration: Consider variable type when picking the aesthetic mapping.

Some aesthetics are ordinal. Some are not. Some can be both!

  • Color palettes, for example, can be:
    • Sequential: Ordered data with one direction
    • Diverging: Ordered data with two directions
    • Qualitative: No order to the data

Consideration: Consider variable type when picking the aesthetic mapping.

Our ability to perceive differences varies by aesthetic!

Consideration: Pick the aesthetic that makes it easy to assess value.

Generally, humans are good at making comparisons involving:

  • Position
  • Length
  • Angle (but less so than position and length!)

Generally, humans are NOT good at making comparisons involving:

  • Area (especially with irregular shapes)
  • Volume
  • Color shade
  • Color hue

Consideration: Respect the principle of proportional ink.

Principle of proportional ink: The size of shaded areas need to be proportional to the data values they represent.

Consideration: Respect the principle of proportional ink.

Principle of proportional ink: The size of shaded areas need to be proportional to the data values they represent.

  • Bars on a linear scale should start at 0.

Consideration: Respect the principle of proportional ink.

Difficult to respect with spatial data. Why?

Consideration: Respect the principle of proportional ink.

Instead of using geographic boundaries, pick a standardized shape and place “near” geographic location.

Consideration: Try to have a high data-ink ratio.

Data-ink ratio: “proportion of a graphic’s ink devoted to the non-redundant display of data-information.” – Edward Tufte

Credit: Mine Çetinkaya-Rundel

Consideration: Think carefully about context.

Consider including:

  • Title (or Figure Caption)
  • Subtitle with maker and data source
  • Caption with key points


  • Legends/helpers (with units)
  • Axis labels (with units)
  • Other annotations or reference points



What to add depends greatly on the research question or the story you are telling!

Consideration: Think carefully about context.

Consider including:

  • Title (or Figure Caption)
  • Subtitle with maker and data source
  • Caption with key points


  • Legends/helpers (with units)
  • Axis labels (with units)
  • Other annotations or reference points



Context should add both memorability and clarity.

Consideration: Simplify as much as you can.

  • Faceting is a great way to add another variable without over-complicating your graphic.

  • But only add additional variables that are useful to the story!

Consideration: Simplify as much as you can.

  • Over-plotting is very common in the Age of Big Data!

  • Example from my own work with the US Forest Inventory and Analysis Program

  • Jitter the points.

Consideration: Simplify as much as you can.

  • Add transparency.

  • Bin the data and then try a different geom.

Pros and cons of different approaches?

Consideration: Make important comparisons easy.

Which graph makes it easy to conclude that the ruling coalition (FDP + SPD) have a majority?

Wilke (2019)

Consideration: Make important comparisons easy.

Which graph makes it easy to see how a company’s market share changes over time? (Warning: Fake data.)

Wilke (2019)

Consideration: Make your graphs accessible!

Not all R color palettes have been vetted for color blindness.

Consideration: Make your graphs accessible!

Not all R color palettes have been vetted for color blindness.

Consideration: Make your graphs accessible!

  • Color contrast also matters.
    • Shoot for a ratio of 4.5 or higher between overlapping colors.

library(coloratio)
cr_get_ratio("#A71F69", "#EBEBEB")
[1] 5.78868
cr_get_ratio("#A4DBE8", "#EBEBEB")
[1] 1.269455

Consideration: Make your graphs accessible!

Use white space to help separate elements.

Consideration: Make your graphs accessible!

Use a large enough font size!

Data Viz Considerations

  • We could spend all semester on data viz principles.
  • Be thoughtful and iterate.
  • Let’s spend some time considering the strengths and weaknesses of some graphs via an activity.
    • Pick out the:
      • Clearest story
      • Most memorable
      • Best overall
  • Be ready to present one of your graphs to the class, if we have time.
15:00