Essentials of Data Visualization

Slide 1

Slide 1 text

Communicating with Data Company A Company B Company C $110 86 62 Sales Name Million Geoff McGhee Bill Lane Center for the American West, Stanford University Behind and Beyond Big Data Worldview Stanford Dec 6, 2016

Slide 2

Slide 2 text

• Why Data Visualization? • How Visualization Works • How to Communicate Data More Effectively • What’s on the Horizon

Slide 3

Slide 3 text

A Decade in Infographics and Multimedia ME

Slide 4

Slide 4 text

ﬂight patterns data art- something in this… but what? http://www.aaronkoblin.com/project/flight-patterns/

Slide 5

Slide 5 text

http://www.bewitched.com/historyflow.html

Slide 6

Slide 6 text

http://hint.fm/projects/flickr/

Slide 7

Slide 7 text

No content

Slide 8

Slide 8 text

• Big data and data visualization are changing news reporting and presentation • New players and skillsets are joining newsrooms • We need to learn more about telling stories with data • Good tools for narrative visualization don’t exist yet Journalism in the Age of Data (2010) datajournalism.stanford.edu

Slide 9

Slide 9 text

No content

Slide 10

Slide 10 text

http://bit.ly/1TS1nQ8

Slide 11

Slide 11 text

Data Visualization Back to Basics

Slide 12

Slide 12 text

No content

Slide 13

Slide 13 text

THEN we drew pictures of data William Playfair, 1786

Slide 14

Slide 14 text

THEN we drew pictures of data NOW we use software code to generate visualizations of (possibly fluctuating) data using predefined rules

Slide 15

Slide 15 text

Visual Encoding of Information But the same fundamental process

Slide 16

Slide 16 text

• “Utilizes one of the channels to our brain that have the highest bandwidths: our eyes”  – Robert Kosara • Bypass language centers, go direct to the visual cortex • Leverage ability to recognize patterns, visual sense-making • Create mental models of phenomena… both literal and metaphorical Map of New Brainland by Unit Seven via Flickr How Visualization Works Why Visualize Information?

Slide 17

Slide 17 text

Literal

Slide 18

Slide 18 text

Fernanda Viegas and Martin Wattenberg

Slide 19

Slide 19 text

Metaphorical

Slide 20

Slide 20 text

The New York Times

Slide 21

Slide 21 text

• Converting – “encoding” – information into graphical marks • Charts, yes, but something more fundamental Lines, shapes, patterns, size, position, shade, hue • The “charts” we know are compounds of those marks How Visualization Works Visual Encoding of Information Company A Company B Company C $110 86 62 Sales Name Million

Slide 22

Slide 22 text

Classical visualizations were essentially done “by seat of the pants” Creators intuited what would work, and it mostly did 20th Century researchers analyzed patterns and put   practices to the test Bertin, Cleveland, Mackinlay et al. How Visualization Works Visual Encoding of Information Sémiologie Graphique  Jacques Bertin, 1967 Graphical Perception  Cleveland and McGill, 1984 Structure of Design Space  Card and Mackinlay, 1997

Slide 23

Slide 23 text

Group Exercise Identify the visual encodings •

Slide 24

Slide 24 text

Linda Eckstein, Fortune

Slide 25

Slide 25 text

John Tomanio, Fortune

Slide 26

Slide 26 text

John Tomanio, Fortune

Slide 27

Slide 27 text

Wittgenstein Center, Vienna Institute for Demography

Slide 28

Slide 28 text

Wittgenstein Center, Vienna Institute for Demography

Slide 29

Slide 29 text

The Wall Street Journal

Slide 30

Slide 30 text

Gregory Hubacek for GOOD Magazine

Slide 31

Slide 31 text

How Visualization Works Visual Encoding of Information The two kinds of graphical perception: Attentive Preattentive

Slide 32

Slide 32 text

How Many Number 5’s? Preattentive Processing

Slide 33

Slide 33 text

Now How Many? Preattentive Processing

Slide 34

Slide 34 text

• We’re much better at quickly detecting SHADE variations than SHAPE differences • First example required reading, because complex shape of a numeral like “5” is an attentive attribute. • Second example: “5’s” are clearly visible when we highlight them in a different shade. Intensity of color is a preattentive attribute. This rapidly occurs below the level of consciousness. • When you find yourself actually reading a chart to try to understand what it says, that’s attentive processing Preattentive Processing How Visualization Works

Slide 35

Slide 35 text

Preattentive Attributes Can be organized into four main categories: Source: Colin Ware: “Information Visualization: Perception for Design” (2004) Preattentive Processing How Visualization Works Color Form Position Movement

Slide 36

Slide 36 text

Not All Preattentive Attributes Behave the Same Way All preattentive attributes can indicate distinctness, but not all have a clear hierarchy that can be utilized for comparing values.  • Quantitative or Ordinal Values:   1, 2, 3... or S, M, L...   Operators =,≠,<,> • Categorical Values:   N, S, E, W... Apples, Oranges  Operators =,≠ > > ? Using Scales to Show Hierarchy How Visualization Works

Slide 37

Slide 37 text

“Quantitative” Visual Attributes Showing X “is greater than” Y Color Intensity Size Line Length Line Thickness Quantitative vs. Qualitative How Visualization Works

Slide 38

Slide 38 text

“Qualitative” or “Categorical” Visual Attributes Showing X “is similar to or different from” Y Hue Shape Enclosure Added Marks Quantitative vs. Qualitative How Visualization Works

Slide 39

Slide 39 text

Pie Angle Size (area)  Hue or intensity (opt.) Bar/Column Size (length of bar) Hue or intensity (opt.) 2D position (neg vals.) Line 2D Position Hue or intensity (for mult. lines) Scatter Plot 2D Position Hue or intensity (opt.) Added marks (icons, etc) Company A Company B Company C $110 86 62 Sales Name Million Text Table Text (yes, text) Numbers 2D Position Added marks (rules, bgs) Bubble Size (area) Hue or intensity 2D Position Network Graph 2D Position Added marks How Visualization Works Charts: Visual Encoding at Work

Slide 40

Slide 40 text

• Position (teams, records, payroll) • Number (record, payroll) • Icons/Text (teams) • Line (connection between record and payroll) • Angle (relationship betw. record and payroll) • Color (relationship between record and payroll) • Line thickness (size of payroll) How Many Encodings? https://fathom.info/salaryper/

Slide 41

Slide 41 text

“I think that there’s a sense that data [visualization] is something more like a medium, something that can be used to tell stories, and to do all of the things that a medium can do, to delight and inspire...” – Eric Rodenbeck, Journalism in the Age of Data (2010)  datajournalism.stanford.edu

Slide 42

Slide 42 text

But...

Slide 43

Slide 43 text

http://bit.ly/286Rmmq

Slide 44

Slide 44 text

http://bit.ly/1NH3q5Q

Slide 45

Slide 45 text

Communication Medium?

Slide 46

Slide 46 text

No content

Slide 47

Slide 47 text

Learning to Tell Stories “You can do beautiful things with computers and lots of data that look very, very nice, and are almost completely incomprehensible.” datajournalism.stanford.edu

Slide 48

Slide 48 text

No content

Slide 49

Slide 49 text

Economist

Slide 50

Slide 50 text

No content

Slide 51

Slide 51 text

No content

Slide 52

Slide 52 text

No content

Slide 53

Slide 53 text

How to Lie with Visualization

Slide 54

Slide 54 text

No content

Slide 55

Slide 55 text

No content

Slide 56

Slide 56 text

No content

Slide 57

Slide 57 text

Example of misleading: Trump bar charts

Slide 58

Slide 58 text

Example of misleading: Trump bar charts

Slide 59

Slide 59 text

Example of misleading: Trump bar charts

Slide 60

Slide 60 text

Example of misleading: Trump bar charts

Slide 61

Slide 61 text

Ways to communicate with data 1. Explain Your Visual Encodings to the User

Slide 62

Slide 62 text

Explain Your Encodings with a Legend, Key, Scale

Slide 63

Slide 63 text

No content

Slide 64

Slide 64 text

Ways to communicate with data 2. Reduce Noise

Slide 65

Slide 65 text

Practice Noise Reduction: The “Data-Ink Ratio” https://darkhorseanalytics.com/blog/data-looks-better-naked/

Slide 66

Slide 66 text

Rainbow palette is popular in scientific visualization But it lacks a visual hierarchy Practice Noise Reduction: Clarity in Symbology

Slide 67

Slide 67 text

Ways to communicate with data 3. Use Narrative Sequences to Tell a Story

Slide 68

Slide 68 text

The “Stepper” Present Main Points of a Visualization in a Sequence http://stanford.io/25zXkdc

Slide 69

Slide 69 text

The “Stepper” Present Main Points of a Visualization in a Sequence http://stanford.io/1G2EopE

Slide 70

Slide 70 text

“Scrollytelling” Present Main Points of a Visualization in a Sequence http://stanford.io/1lZ1vXO

Slide 71

Slide 71 text

“Scrollytelling” Present Main Points of a Visualization in a Sequence http://stanford.io/1qppbav

Slide 72

Slide 72 text

Other Narrative Formats Simulators and Models http://stanford.io/1lZ1vXO

Slide 73

Slide 73 text

Other Narrative Formats Simulators and Models http://projects.aljazeera.com/2013/syrias-refugees/index.html

Slide 74

Slide 74 text

Ways to communicate with data 4. Annotate Liberally to Explain What’s Shown

Slide 75

Slide 75 text

• As powerful as visualization can be to enable rapid sense-making, many visualizations require a lot of context and interpretation, especially static, printed visualizations. • Think of the annotation layer as like the narrator’s voiceover in a documentary film. In fact, if you make a motion video, that might literally what it would be. Raise Your Narrative Voice The New York Times The Annotation Layer

Slide 76

Slide 76 text

http://nyti.ms/1qfwceV

Slide 77

Slide 77 text

http://nyti.ms/1qfwceV

Slide 78

Slide 78 text

Ways to communicate with data 5. Where Possible, Offer Record-Level Detail

Slide 79

Slide 79 text

• Also known as “details on demand,” interactive visualizations are richer when they give you as much info as possible on individual measures • In that sense, they are database front-ends in addition to graphical summaries. Visualizations are Databases Where Possible, Offer Record-Level Detail

Slide 80

Slide 80 text

Where Possible, Offer Record-Level Detail

Slide 81

Slide 81 text

http://citynature.stanford.edu/ Where Possible, Offer Record-Level Detail

Slide 82

Slide 82 text

http://citynature.stanford.edu/ Where Possible, Offer Record-Level Detail

Slide 83

Slide 83 text

Ways to communicate with data 1. Explain Your Symbology/Visual Encodings to the User 2. Reduce Noise 3. Use Narrative Sequences to Tell a Story 4. Annotate Liberally to Explain What’s Shown 5. Where Possible, Offer Record-Level Detail