EDS 240 – lecture2.1-choosing-graphic-forms-slides

EDS 240: Lecture 2.1

Choosing the right graphic form

Week 2 | January 13^th, 2024

We understand complex numbers better when they’re represented visually

“Exams will have a total of 137 points rather than the usual 100. This scoring system has no effect on the grade you get in the course, but it seems to make you happier”

-Richard H. Thaler, economist & professor

Early years: exam graded 0 - 100 with an average score of 72 points = lots of complaints
Later years: exam graded 0 - 137 with an average score of 96 points = very few complaints

Albert Cairo’s visualization of the scores from Thaler’s exam case study

Vision is our most well-developed sense

Mapping data into visual properties is powerful

Mapping?

How values of a variable(s) of interest are represented by visuals (e.g height of bar, shaded region of area plot, color of data points)

How do you choose the right graphic form to represent your data?

“If I had the answer to that, I’d be rich by now…I have no idea, but I can give you some clues to make your own choices based on what we know about why and how visualization works”

-Albert Cairo¹, in his book, The Truthful Art

Exercise: Map data to visual properties

03:00

Let’s say you want to compare unemployment rates of 5 countries: A, B, C, D, E (the actual values here are not important).

How would you map the unemployment rates to visual properties in a way that enables your readers to accurately compare values without having to read all the numbers?

Different methods of encoding the same data

“Hierarchy of elementary perceptual tasks”

William S. Cleveland & Robert McGill (1984) Graphical Perception: Theory, Experimentation, and Application to the Development of Graphical Methods, Journal of the American Statistical Association, 79:387, 531-554, DOI: 10.1080/01621459.1984.10478080

a viewer performs one or more of these mental-visual tasks (judging position, perceiving angles / areas, etc.) to extract the values of real variables represented on most graphs
successful charts are constructed based on elementary tasks “as high in the hierarchy as possible”

Albert Cairo’s recreation of Cleveland & McGill’s Hierarchy of Elementary Perceptual Tasks

Exercise: How many times bigger is the larger circle?

01:00

Exercise: How many times bigger is the larger bar?

01:00

Both the circles & rectangles differ by a magnitude of 7

Caveats to the hierarchy

1. Cleveland & McGill only considered statistical charts. What about data maps, for example, that rely on area / shading / hue, which fall lower on the hierarchy?

Caveats to the hierarchy - an example

Lower scale methods can be appropriate when the goal is to reveal general patterns. For example, a choropleth map displays divided geographical areas / regions, which are colored in relation to a numeric variable.

Precipitation data downloaded from NOAA’s National Centers for Environmental Information.

Caveats to the hierarchy

1. Cleveland & McGill only considered statistical charts. What about data maps, for example, that reply on area / shading / hue, which fall lower on the hierarchy?

2. No method of choosing a graphic form is perfect! It’s important to think critically about your graphic’s purpose and how best to represent your data to serve that purpose.

Caveats to the hierarchy - an example

Consider how you might display the same data in the following Sankey diagram, which depicts the flow of refugees in 2022, using graph types from the top of Cleveland & McGill’s hierarchy. What is the purpose of this chart?

Graphic recreated using the {networkD3} package following Louise E. Sinks’ blog post, TidyTuesday: Exploring Refugee Flow with A Sankey Diagram

Tips for choosing the right graphic form

Think about the task(s) you want to enable or message(s) you want to convey. For example, do you want to compare, see change or flow, reveal relationships or connections, envision temporal or spatial patterns.

Consider the number of variables and the number of data points, as well as the data types you’re working with. For example, do you have several vs. many data points? How many categorical and/or numeric variables? Are your variables ordered or not ordered? Data types can dictate which graphical form is appropriate.

Try different graphic forms, especially if you have more than one task to enable or message to convey.

Arrange the components of the graphic to make it as easy as possible to extract meaning from your graphic quickly.

Test the outcomes of your graphic on others, particularly on those who are representative of the audience you are trying to reach.

Tips for choosing the right graphic form

Think about the task(s) you want to enable or message(s) you want to convey. For example, do you want to compare, see change or flow, reveal relationships or connections, envision temporal or spatial patterns.
Consider the number of variables and the number of data points, as well as the data types you’re working with. For example, do you have several vs. many data points? How many categorical and / or numeric variables? Are your variables ordered or not ordered? Data types can dictate which graphical form is appropriate.
Try different graphic forms, especially if you have more than one task to enable or message to convey.
Arrange the components of the graphic to make it as easy as possible to extract meaning from your graphic quickly.
Test the outcomes of your graphic on others, particularly on those who are representative of the audience you are trying to reach.

1. What task(s) to enable / message(s) to convey

2. Number of variables & data points, data types

Data Viz Project displays one small data set 100 different ways

From Data to Viz search graphic types by data type or by function (+ R & Python Graph Gallery)

The Visualization Universe compares most popular graphic forms

02:00

2. Number of variables & data points, data types

Quantitative data

Cartoon comparison of continuous versus discrete data. On the left: 'Continuous - measured data, can have infinite values within possible range.' Below is an illustration of a chick, with text 'I am 3.1 inches tall, I weight 34.16 grams.' On the right: 'Discrete - observations can only exist at limited values, often counts.' Below is an illustration of an octopus with text 'I have 8 legs and 4 spots!'

Continuous variables: temperature (10.6°C, 14.9°C, 8.1°C), rainfall (1.7”, 3.3”, 9.4”)

Discrete variables: # of species counted in a region (1, 4, 6), a county’s population size (1,578, 10,324, 540,013)

2. Number of variables & data points, data types

Quantitative data

Continuous variables: temperature (10.6°C, 14.9°C, 8.1°C), rainfall (1.7”, 3.3”, 9.4”)

Discrete variables: # of species counted in a region (1, 4, 6), a county’s population size (1,578, 10,324, 540,013)

Qualitative data

Visual representations of nominal, ordinal, and binary variables. Left: Nominal (ordered descriptions) with illustrations below of a turtle, snail, and butterfly. Center: Ordinal (ordered descriptions) with illustrations below of three bees - one looks unhappy (saying 'I am unhappy'), one looks ok (saying 'I am OK'), and one looks very happy (saying 'I am awesome!'). Right: Binary (only 2 mutually exclusive outcomes), with below a T-rex saying 'I am extinct' and a shark saying 'HA.'

Nominal variables: gender identity (cisgender, transgender, non-binary), species (dog, cat, bird), land use (residential, parks, agriculture)

Ordinal variables: income level (low / middle / high), satisfaction level (unsatisfied, neutral, satisfied)

Binary: penguin sex (male / female), habitat type (shade / sun)

Tips for choosing the right graphic form

Think about the task(s) you want to enable or message(s) you want to convey. For example, do you want to compare, see change or flow, reveal relationships or connections, envision temporal or spatial patterns.
Consider the number of variables and the number of data points, as well as the data types you’re working with. For example, do you have several vs. many data points? How many categorical and/or numeric variables? Are your variables ordered or not ordered? Data types can dictate which graphical form is appropriate.
Try different graphic forms especially if you have more than one task to enable or message to convey.
Arrange the components of the graphic to make it as easy as possible to extract meaning from your graphic quickly.
Test the outcomes of your graphic on others, particularly on those who are representative of the audience you are trying to reach.

3. Try different graphic forms

If we want to show both big picture patterns and detailed comparisons, we may consider including multiple graphic forms in the same visualization.

Our choropleth map displays 5-year precipitation as compared to the 20th century average by US county. But what if we wanted to explore precipitation across CA counties? A choropleth, which uses colors to encode information, may not be the easiest for making these comparisons. Choosing a graphical form from the top of the hierarchy (e.g. bar chart) may be more effective.

Tips for choosing the right graphic form

Think about the task(s) you want to enable or message(s) you want to convey. For example, do you want to compare, see change or flow, reveal relationships or connections, envision temporal or spatial patterns.
Consider the number of variables and the number of data points, as well as the data types you’re working with. For example, do you have several vs. many data points? How many categorical and/or numeric variables? Are your variables ordered or not ordered? Data types can dictate which graphical form is appropriate.
Try different graphic forms, especially if you have more than one task to enable or message to convey.
Arrange the components of the graphic to make it as easy as possible to extract meaning from your graphic quickly.
Test the outcomes of your graphic on others, particularly on those who are representative of the audience you are trying to reach.