Survival Skills: Concepts of Data-Driven Visualization

University of Oklahoma Libraries

Getting started

What we will discuss today

  • Define visualizations for data

  • What, why, and how of data-driven visualizations

    • What: goals of graphing

    • Why: human perception

    • How: design and defaults

  • How to get more help

Ice breakers

  • 5 min ice breaker activity: A boring fact about yourself

What is data-driven visualization?

  • Definition: “visual representation and presentation of data to facilitate understanding”
  • We use visualizations (graphs, figures, plots) for
    • exploring
    • telling a story

Consider this visualization (5 min)

  • What variables are included?

  • What question could this graph answer?

  • What is confusing to you?

Again, here’s what we’ll cover about visualizations

  • What, why, and how of data-driven visualizations

    • What: goals

    • Why: human perception

    • How: design and defaults

  • Lots of new and classic visualization resources!

Handout (easy print PDF)

Handout (clickable HTML)

Goals: What you need the graph to be

Trustworthy, reasonable, faithful, integrity1

  • Ensure all variables from question are included, aka data density2
  • Are the relevant summaries and uncertainty shown?3

Where appropriate, follow examples

  • What’s the common way to present this in your discipline?
    • Searching in captions in some databases.
  • Check your journal style guide if submitting manuscript

Possible + useful = relevant1

  • There is no right answer, there is the right answer for your situation
  • Graph data more than once to show different aspects

Perception: Why graphs work

But first, your audience/viewer

  • Exhausted
  • Busy
  • Indifferent

“How your email finds me”

Presentation, accessible, usable, understandable1

  • Visual patterns as an augment to working memory limitations of humans2
     x   y z
1  0.0 0.5 c
2  2.0 1.0 c
3  3.0 1.0 c
4  4.0 2.0 c
5  6.0 1.5 c
6  2.5 2.0 t
7  3.0 3.0 t
8  3.5 4.0 t
9  4.0 5.0 t
10 6.0 6.0 t

Limits of “typical” human perception

  • Universal design
    • Size of visualization elements (text, symbols, lines)
  • Perceptual accuracy and “ordered elementary tasks”1

Top three “perceptually accurate representations”

  1. Position along a common scale (common baseline)

Top three “perceptually accurate representations”

  1. Position along identical, non-aligned scales (small multiples1)

Top three “perceptually accurate representations”

  1. Length - both figures have a 1 “unit” difference

More difficult to judge1

  • Angle/slope
    • We can do slopes, but not angles (this is part of why pie charts difficult)
  • Area
  • Volume
  • Color hue - color saturation - density

When in doubt

  • Choose the principle least likely to mislead1
  • Use more than one aspect for redundancy
    • For example, use both shape and color

Design: How to communicate your message

Consider location of audience

  • Paper, screen, projected
  • Not all minds or eyes work the same - universal design
  • Make changes to tell your story more clearly1
    • Program defaults usually not great
    • Learn how to save settings for graphs in code

What actions to take?

  • Three ways to design systematically1

    • Highlight
    • Organize
    • “Integrate text” (context)

Highlight

Show what you need to show!

Highlight: make data obvious

  • Make data visually prominent1

Highlight: effective differences

Highlight: effective differences

  • Maximum contrast/difference - does it help?

Highlight: effective differences

Highlight: effective differences

  • Use smallest1 needed contrast

Highlight: calculate before plotting

  • Do the work for the viewer1
    • This means plot the variable of interest
    • Differences, not before and after

Organize

“Visual complexity is distracting and should therefore never be employed to a degree that exceeds the actual complexity in the data”1

Organize: comparisons

  • Use common baselines wherever possible to make group comparisons1

Organize: separate overlaps

  • Use small multiples to show otherwise overlapping groups1

Organize: reduce clutter

  • Reduce interior clutter1 in grids, ticks, labels

Integrate with context

Context: consistency with text

  • Label variables consistent with text1, not data abbreviations
     x   y z
1  0.0 0.5 c
2  2.0 1.0 c
3  3.0 1.0 c
4  4.0 2.0 c
5  6.0 1.5 c
6  2.5 2.0 t
7  3.0 3.0 t
8  3.5 4.0 t
9  4.0 5.0 t
10 6.0 6.0 t

Context: consistency in symbols (default, not good)

  • Use consistent symbology in related visualizations1

Context: consistency in symbols (updated)

  • Use consistent symbology in related visualizations1

Context: labelling

  • Label directly on visuals where possible1

Ending activity

Think about the original graph

  • What variables are included?

    • SPL dB(Z)
    • Frequency (Hz)
    • Some sort of heat map / intensity
    • Types of experiments/treatments
    • Annotation highlights

  • What is confusing?

    • What are annotations?
    • Treatment combinations
    • x-axis ticks not proportional

Some possibilities

  • Goals
    • Effect size not reported
  • Highlight
    • Make lines thicker, add shapes at data
    • Highlight relevant frequencies
  • Organize
    • Small multiples for effect sizes
    • Make comparisons clearer with heat map
  • Context
    • Label variables
    • Make proportionally accurate x-axis
    • Integrate with text via images

Get more help

OU Libraries Digital Scholarship and Data Services

  • Open office hours (drop in)
  • Schedule an appointment with our contact form

libraries.ou.edu/data

Please provide feedback

ou.libwizard.com/f/datasurvey