[PPT] - Best Practices in Data Visualization Jodie Jenkinson, Associate PowerPoint Presentation

SLIDE 1

Best Practices in Data Visualization

Jodie Jenkinson, Associate Professor + Director Biomedical Communications University of Toronto

bmc.med.utoronto.ca

Best Practices in Data Visualization – ComSciComCan Jenkinson

Because of the inexplicability of complex information or raw data
In order to leverage visual perception
To create an aid to understanding
To provide insight

Why visualize data?

SLIDE 2

Best Practices in Data Visualization – ComSciComCan Jenkinson

40%+ of cortex devoted to visual perception
High-bandwidth channel
Parallel processing

Leveraging visual perception

Best Practices in Data Visualization – ComSciComCan Jenkinson

Information processing capacity of the visual system
109 bits per second*
= 1 billion bits
= ~120 Megabytes per second

Vision is high bandwidth

* Information Capacity of a Single Retinal Channel, DH Kelly, IRE Transactions on Information Theory, 1962, pp. 221

All visual stimuli Pre-attentive Pre-attentive Attend to

SLIDE 3

Best Practices in Data Visualization – ComSciComCan Jenkinson

40%+ of cortex devoted to visual perception
High-bandwidth channel
Parallel processing

Leveraging visual perception

Best Practices in Data Visualization – ComSciComCan Jenkinson

Information processing capacity of the visual system
109 bits per second*
= 1 billion bits
= ~120 Megabytes per second

Vision is high bandwidth

* Information Capacity of a Single Retinal Channel, DH Kelly, IRE Transactions on Information Theory, 1962, pp. 221

All visual stimuli Pre-attentive Pre-attentive Attend to

SLIDE 4

Best Practices in Data Visualization – ComSciConCan Jenkinson

Providing insight

I II III IV x y x y x y x y 10 8.04 10 9.14 10 7.46 8 6.58 8 6.95 8 8.14 8 6.77 8 5.76 13 7.58 13 8.74 13 12.74 8 7.71 9 8.81 9 8.77 9 7.11 8 8.84 11 8.33 11 9.26 11 7.81 8 8.47 14 9.96 14 8.1 14 8.84 8 7.04 6 7.24 6 6.13 6 6.08 8 5.25 4 4.26 4 3.1 4 5.39 19 12.5 12 10.84 12 9.13 12 8.15 8 5.56 7 4.82 7 7.26 7 6.42 8 7.91 5 5.68 5 4.74 5 5.73 8 6.89

Anscombe’s quartet

Best Practices in Data Visualization – ComSciComCan Jenkinson

Same
Mean (x and y)
Variance
Correlation
Regression

Why visualize?

Anscombe’s quartet

I II III IV x y x y x y x y 10 8.04 10 9.14 10 7.46 8 6.58 8 6.95 8 8.14 8 6.77 8 5.76 13 7.58 13 8.74 13 12.74 8 7.71 9 8.81 9 8.77 9 7.11 8 8.84 11 8.33 11 9.26 11 7.81 8 8.47 14 9.96 14 8.1 14 8.84 8 7.04 6 7.24 6 6.13 6 6.08 8 5.25 4 4.26 4 3.1 4 5.39 19 12.5 12 10.84 12 9.13 12 8.15 8 5.56 7 4.82 7 7.26 7 6.42 8 7.91 5 5.68 5 4.74 5 5.73 8 6.89

SLIDE 5

Best Practices in Data Visualization – ComSciConCan Jenkinson

Why visualize?

I II III IV x y x y x y x y 10 8.04 10 9.14 10 7.46 8 6.58 8 6.95 8 8.14 8 6.77 8 5.76 13 7.58 13 8.74 13 12.74 8 7.71 9 8.81 9 8.77 9 7.11 8 8.84 11 8.33 11 9.26 11 7.81 8 8.47 14 9.96 14 8.1 14 8.84 8 7.04 6 7.24 6 6.13 6 6.08 8 5.25 4 4.26 4 3.1 4 5.39 19 12.5 12 10.84 12 9.13 12 8.15 8 5.56 7 4.82 7 7.26 7 6.42 8 7.91 5 5.68 5 4.74 5 5.73 8 6.89

Anscombe’s quartet

4 6 9 11 13 4 8 12 16 20 4 6 9 11 13 4 8 12 16 20 4 6 9 11 13 4 8 12 16 20 4 6 9 11 13 4 8 12 16 20 Francis J. Anscombe, Graphs in Statistical Analysis. The American Statistician, vol. 27, no. 1, pp. 17–21, 1973

“…If a picture is only worth a thousand words, we're screwed.”

Eric Lander Professor of Biology, MIT VIZBI 2011, Opening Remarks

Best Practices in Data Visualization – ComSciComCan Jenkinson

Visual mapping (encoding data)
Providing adequate context
Balancing clarity & aesthetics

Data Visualization in a nutshell Encoding Data

SLIDE 6

Best Practices in Data Visualization – ComSciComCan Jenkinson

Visual representation of data should be consistent with the numerical

representation

Title Text

Best Practices in Data Visualization – ComSciComCan Jenkinson

Nominal
name, type, category
eg. mammals, reptiles, birds
Ordinal
integer sequence
eg. first, second, third
happy, very happy, ecstatic

Data types

Best Practices in Data Visualization – ComSciComCan Jenkinson

Interval
gap in values
eg. every three months
Ratio
real numbers; zero as reference
45.7 out of 100

Data types

Best Practices in Data Visualization – ComSciComCan Jenkinson

Spatial
eg. maps, GIS, directions
scalar fields
Narrative
eg. assembly sequence, process

Data features

SLIDE 7

Best Practices in Data Visualization – ComSciConCan Jenkinson

Data features

1d 2d 3d

Best Practices in Data Visualization – ComSciComCan Jenkinson

The way in which data is mapped to visual structures
Every visualization can be described as a set of mappings:
from data items to visual marks
from data attributes to visual channels

Visual encoding

Best Practices in Data Visualization – ComSciComCan Jenkinson

Data Items:
Data Marks: the basic visual units that represent data objects visually
Data Attributes:
Visual Channels: the visual variables we can use to represent

characteristics of these objects

Title Text

Best Practices in Data Visualization – ComSciConCan Jenkinson

Marks

From Enrico Bertini

SLIDE 8

Best Practices in Data Visualization – ComSciConCan Jenkinson

Channels

From Enrico Bertini Best Practices in Data Visualization – ComSciConCan Jenkinson

Cleveland & McGill’s Perceptual Task Scale

Allows more accurate judgments Allows more generic judgments Position along a common scale Position along nonaligned scales Length Colour saturation Shading Curvature Volume Area Angle Direction

From The Functional Art, Alberto Cairo Best Practices in Data Visualization – ComSciComCan Jenkinson

Tasks are grouped according to how well you can perceive differences in

the data

Cleveland and McGill’s Perceptual Task Scale

Best Practices in Data Visualization – ComSciComCan Jenkinson

People are not good at making visual angular distinctions
Pie charts are sometimes rolled out to encode 1 or 2 numbers; usually a

very low data density!

A basic example: the pie chart

SLIDE 9

Best Practices in Data Visualization – ComSciConCan Jenkinson

Pie charts

2002 2003 2004 2005 2006 2007

Best Practices in Data Visualization – ComSciConCan Jenkinson

Pie charts

7% 8% 10% 11% 29% 35%

2002 2003 2004 2005 2006 2007

“There are three kinds of lies: lies, damned lies, and statistics”

Benjamin Disraeli

Best Practices in Data Visualization – ComSciComCan Jenkinson

Misalignments between graphic elements and the data they are

intended to represent

Lies, damned lies…

SLIDE 10

Best Practices in Data Visualization – ComSciConCan Jenkinson

Misleading use of area

Best Practices in Data Visualization – ComSciConCan Jenkinson

Misleading use of area

Best Practices in Data Visualization – ComSciConCan Jenkinson

Misleading use of area

Best Practices in Data Visualization – ComSciConCan Jenkinson

Misleading use of area…

SLIDE 11

Best Practices in Data Visualization – ComSciComCan Jenkinson

A common mistake for chart design is to scale an area by two sides at

the same time, producing a quadratic effect for a linear change

Linear vs. quadratic change

Tie area of the white square = a

2

a = 80 Tie area of the white square =6,400 px To double the area of white square =12,800 px a = 160 Tie area of the greysquare a = 25,600 px

2

...four times that of the white square Tie area of red square is twice that of white square √12,800 = 113 a = 113

Best Practices in Data Visualization – ComSciConCan Jenkinson

Linear vs quadratic change

The area of the blue circle is equal to πr2 (20,106) The area of the red circle is equal to πr2 (80, 424)

r = 80 r = 160

Best Practices in Data Visualization – ComSciConCan Jenkinson

Size encoding

Height Area Volume

Best Practices in Data Visualization – ComSciComCan Jenkinson

Modifications to the X or Y axis in an attempt to make differences or

change appear to be more dramatic

Data represented out of context does not allow for adequate comparison

How statistics lie

SLIDE 12

Best Practices in Data Visualization – ComSciConCan Jenkinson

The disappearing baseline

Best Practices in Data Visualization – ComSciConCan Jenkinson

The disappearing baseline

Best Practices in Data Visualization – ComSciConCan Jenkinson

Reversing the x-axis

SLIDE 13

Best Practices in Data Visualization – ComSciConCan Jenkinson

Reversing the x-axis

Best Practices in Data Visualization – ComSciConCan Jenkinson

No defined y-axis

Best Practices in Data Visualization – ComSciConCan Jenkinson

No defined y-axis Context

SLIDE 14

Best Practices in Data Visualization – ComSciComCan Jenkinson

Without context we are unable to see the big picture
Without context we are unable to make meaningful comparisons

Context

Source: New York Times Source: New York Times Source: New York Times

SLIDE 15

Best Practices in Data Visualization – ComSciConCan Jenkinson

Data in and out of context

Best Practices in Data Visualization – ComSciConCan Jenkinson

Data in and out of context

Best Practices in Data Visualization – ComSciConCan Jenkinson

Data in and out of context

Best Practices in Data Visualization – ComSciConCan Jenkinson

Historical context

SLIDE 16

Best Practices in Data Visualization – ComSciConCan Jenkinson

Historical context Clarity & Aesthetics

Best Practices in Data Visualization – ComSciComCan Jenkinson

The important of the information should match the salience of the

channel

Effectiveness

Kim OY, et al. (2012) Higher levels of serum triglyceride and dietary carbohydrate intake are associated with smaller LDL particle size in healthy Korean women. Nutrition Research and Practice 6:120-125

SLIDE 17

Cawley S, et al. (2004) Unbiased mapping of transcription factor binding sites along human chromosomes 21 and 22 points to widespread regulation of noncoding RNAs. Cell 116:499-509

French Polynesia Chad Afghanistan Congo (Kinshasa) Burundi Cambodia Uganda Mali Ethiopia Malawi Burkina Faso Central African Republic Somalia Rwanda Laos Niger Nepal Lesotho Tanzania Madagascar Guinea Eritrea Comoros Liberia Gambia, The Sierra Leone Haiti Zambia Mozambique Burma(Myanmar) Guinea-Bissau Bangladesh Kenya Sudan Ghana Benin Cote d'Ivoire (IvoryCoast) Kiribati Solomon Islands Cameroon Togo Senegal Bhutan Vanuatu Sao Tome and Principe Paraguay Sri Lanka Cape Verde Turks and Caicos Islands Nigeria Pakistan Samoa Philippines Papua New Guinea Nicaragua

W. Sahara

Yemen Zimbabwe Guatemala Swaziland El Salvador Kyrgyzstan Mauritania Georgia Honduras Morocco Peru Tajikistan Vietnam U.S. Pacific Islands India Indonesia Albania Namibia Tonga Dominica Bolivia Costa Rica Colombia Fiji Congo (Brazzaville) Saint Vincent/Grenadines Moldova Angola Saint Helena Uruguay Ecuador Dominican Republic Egypt Brazil Maldives Tunisia Saint Lucia Botswana Cuba Syria Algeria Mongolia Grenada Gabon Mauritius Cook Islands Turkey North Korea Jordan Belize Armenia Macedonia Saint Kitt and Nevis Thailand Latvia Bosnia and Herzegovina Chile Mexico Argentina Djibouti Virgin Islands British Suriname Jamaica Lithuania Romania Uzbekistan Panama China Croatia Former Serbia and Montenegro Azerbaijan Macau Barbados Portugal Hungary Venezuela Switzer- land Sweden France Bulgaria Belarus Malaysia Slovakia Ukraine Iran Montserrat Malta Poland Italy Equatorial Guinea Slovenia Libya Cayman Islands Spain Austria New Zealand Aruba Antigua and Barbuda United Kingdom Japan Norway Israel Greece Turkmenistan South Africa Bermuda Germany Korea, South Puerto Rico American Samoa Denmark Greenland Guam Finland Oman Seychelles Czech Republic Cyprus Iceland Ireland New Caledonia Saint Pierre and Miquelon Russia Hong Kong Taiwan Nauru Kazakhstan Estonia Belgium Faroe Islands Saudi Arabia Netherlands The Bahamas Canada United States Australia Luxembourg Brunei Kuwait

Singapore

United Arab Emirates Bahrain Trinidad and Tobago

Netherlands Antilles Virgin Islands, U.S. Gibraltar

Iraq Lebanon

Qatar

Serbia Nepal Chad Wake Island U.S. Pacific Islands Timor-Leste (East Timor) Solomon Islands Papua New Guinea Maldives Laos French Polynesia Cook Islands Brunei Bhutan New Zealand American Samoa Afghanistan Cote d’Ivoire Western Sahara Tanzania Seychelles Sao Tome and Principe Saint Helena Guinea-Bissau Madagascar Mozambique Equatorial Guinea Congo (Brazzaville) Comoros Central African Republic Cape Verde Burkina Faso Yemen Palestine Oman Lebanon Bahrain Tajikistan Moldova Kyrgyzstan Estonia Azerbaijan Armenia Macedonia Iceland Hungary Gibraltar Former Serbia and Montenegro Faroe Islands Cyprus Bosnia and Herzegovina Albania Dominica Virgin Islands, U.S. Turks and Caicos Islands Trinidad and Tobago Suriname Saint Vincent/Grenadines Saint Lucia Saint Kitts and Nevis Guatemala Nicaragua Netherlands Antilles Martinique Guadeloupe French Guiana Falkland Islands (Islas Malvinas) El Salvador Jamaica Costa Rica Cayman Islands Sierra Leone Paraguay Bahamas, The Saint Pierre and Miquelon Antigua and Barbuda Greenland Bermuda New Caledonia Virgin Islands, British Tonga Timor-Leste (East Timor) Tonga Samoa Nauru Uganda Tunisia Somalia Senegal Rwanda Niger Liberia Lesotho Guinea Gabon Eritrea Burundi Malta Haiti Panama Grenada Guyana Belize Honduras Aruba Zambia Togo Swaziland Namibia Morocco Mauritius Mauritania Mali Malawi Kenya Ghana Gambia Ethiopia Djibouti Cameroon Botswana Benin Angola Macau Cambodia Zimbabwe Turkmenistan Mongolia Luxembourg North Korea Vietnam Thailand

Taiwan

Singapore Philippines Pakistan Malaysia

South Korea Japan

Indonesia

India

Hong Kong

China

Sri Lanka Burma Bangladesh Fiji Guam

Australia

Reunion Sudan

South Africa Egypt

Congo (Kinshasa) Algeria United Arab Emirates Syria Saudi Arabia Qatar Libya Nigeria Kuwait Jordan Israel Iraq Iran Uzbekistan

Ukraine Russia

Lithuania Kazakhstan Georgia Belarus United Kingdom

Turkey

Switzerland Sweden

Spain

Slovakia Romania Portugal Poland Norway Netherlands Slovenia

Italy

Greece

Germany France

Finland Denmark Czech Republic Croatia Bulgaria Belgium Austria Venezuela Chile Peru Ecuador Colombia Brazil Bolivia Montserrat Uruguay Argentina Cuba Barbados Puerto Rico

United States

Dominican Republic

Mexico

Latvia Ireland

Canada

EUROPE

Total Carbon Emissions by Nation

Tracking Carbon Emissions

KEY

Per Capita Carbon Emissions by Nation

ASIA MIDDLE EAST AFRICA NORTH AMERICA CENTRAL AMERICA SOUTH AMERICA OCEANIA CARIBBEAN NOTE: BASED ON 2007 DATA. SOURCES: U.S. ENERGY INFORMATION ADMINISTRATION DESIGN: STANFORD KAY STUDIO.COM China’s total emissions lead the world, but when diluted by its huge population, its ranking drops down the per capita list. Guyana The United States is

no. 2 for total emissions

but Americans shrink down to a respectable rank in line with other industrialized citizens. Tiny Gibraltar tops the per capita list due to its need to import most manufactured goods— a reality also seen in many small island nations.

A footprint comparison of total carbon dioxide emissions by nation and per capita shows there’s plenty of room for smaller countries to reduce their carbon footprints. By Stanford Kay

°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°

“The data-ink ratio is the proportion of ink that is used to present actual data compared to the to the total amount of ink used in the entire display.”

Edward Tufte

SLIDE 18

Data/Ink = ink used for data total ink used in graphic

“The data-ink ratio is the proportion of ink that is used to present actual data compared to the to the total amount of ink used in the entire display.”

Best Practices in Data Visualization – ComSciComCan Jenkinson

Data arranged in columns and rows
Data encoded as words and numbers

Anatomy of a table

Time Topic Lecturer 2:10 pm The fine art of torture Ramsay Bolton 3:00 pm Break

3:10pm

Lessons in servitude Theon Greyjoy

Best Practices in Data Visualization – ComSciComCan Jenkinson

When “looking up” is important
When precision and value comparison are required
When units vary

When to use tables

Mortgage Rate Bank Type 5.2% EDQ Fixed 5% BMA Variable 4.8% INF Variable

Best Practices in Data Visualization – ComSciComCan Jenkinson

Are the grids necessary?

Design guidelines

Mortgage Rate Bank Type 5.2% EDQ Fixed 5% BMA Variable 4.8% INF Variable

SLIDE 19

Best Practices in Data Visualization – ComSciComCan Jenkinson

Usually, no

Design guidelines

Mortgage Rate Bank Type 5.2% EDQ Fixed 5% BMA Variable 4.8% INF Variable

Best Practices in Data Visualization – ComSciComCan Jenkinson

Wide tables sometimes need a graphic device to aid horizontal scanning

Design guidelines

Train Destination Time 23 Poughkeepsie 1:00 2:00 3:00 4:00 5:00 6:00 7:00 48 Timbuctoo 1:15 2:15 3:15 4:15 5:15 6:15 7:15 15 Wawa 1:30 2:30 3:30 4:30 5:30 6:30 7:30

Best Practices in Data Visualization – ComSciComCan Jenkinson

Wide tables sometimes need a graphic device to aid horizontal scanning

Design guidelines

Train Destination Time 23 Poughkeepsie 1:00 2:00 3:00 4:00 5:00 6:00 7:00 48 Timbuctoo 1:15 2:15 3:15 4:15 5:15 6:15 7:15 15 Wawa 1:30 2:30 3:30 4:30 5:30 6:30 7:30

Best Practices in Data Visualization – ComSciComCan Jenkinson

Wide tables sometimes need a graphic device to aid horizontal scanning

Design guidelines

Train Destination Time 23 Poughkeepsie 1:00 2:00 3:00 4:00 5:00 6:00 7:00 48 Timbuctoo 1:15 2:15 3:15 4:15 5:15 6:15 7:15 15 Wawa 1:30 2:30 3:30 4:30 5:30 6:30 7:30

SLIDE 20

Best Practices in Data Visualization – ComSciConCan Jenkinson

Milton GO Bus Service

Toronto Union->Mississauga->Meadowvale-> Milton zn Table Monday to Friday (excluding holidays) 2 Toronto-Union Bus Term 07 20 08 20 09 20 10 20 10 50 11 20 12 00 12 10 12 20 2 Union Station ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ 3 Kipling GO Station ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ 11 Dixie GO Station 07 40 08 40 09 40 10 40 11 10 11 40 12 20 ¯ ¯ 11 Cooksville GO Station 07 50 08 50 09 50 10 50 11 20 11 50 12 30 ¯ ¯ 20 Square One-Rathbum@D.O.Y. 08 00 09 00 10 00 11 00 11 30 12 00 12 40 ¯ ¯ 12 Erindale GO Station 08 05 09 05 10 05 11 05 11 35 12 05 12 40 ¯ 21 Streetsville GO Station 09 15 10 15 11 15 12 15 12 50 ¯

Best Practices in Data Visualization – ComSciConCan Jenkinson

Milton GO Bus Service

Toronto Union ➔ Mississauga ➔ Meadowvale ➔ Milton

Zone & Station Monday to Friday (excluding holidays) 02 Toronto-Union Bus Term 07:20 08:20 09:20 10:20 10:50 11:20 12:00 12:10 12:20 02 Union Station ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ 03 Kipling GO Station ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ 11 Dixie GO Station 07:40 08:40 09:40 10:40 11:10 11:40 12:20 ¯ ¯ 11 Cooksville GO Station 07:50 08:50 09:50 10:50 11:20 11:50 12:30 ¯ ¯ 20 Square One-Rathburn@D.O.Y. 08:00 09:00 10:00 11:00 11:30 12:00 12:40 ¯ ¯ 12 Erindale GO Station 08:05 09:05 10:05 11:05 11:35 12:05 12:40 ¯ 21 Streetsville GO Station 09:15 10:15 11:15 12:15 12:50 ¯

Best Practices in Data Visualization – ComSciComCan Jenkinson

Our ability to perceive colour is surprisingly limited
Much of the information we perceive is derived from luminance (light and

shadow)

Colour guidelines

SLIDE 21

Best Practices in Data Visualization – ComSciComCan Jenkinson

To attract attention
To group elements
To indicate meaning
To enhance aesthetics

Fundamental uses of colour

Best Practices in Data Visualization – ComSciComCan Jenkinson

Reserve saturation for areas requiring attention
Background colours should be low in saturation
If you don’t need colour don’t use it

Colour recommendations

SLIDE 22

Best Practices in Data Visualization – ComSciComCan Jenkinson

Harmonizing text and line-drawing requires sensitive appraisals of

interaction effects

Unless necessary, avoid boxing in text as this activates negative space

between the word and the boxes

Typography guidelines

Best Practices in Data Visualization – ComSciComCan Jenkinson

UPPER CASE TYPE IS MORE DIFFICULT TO READ THAN lower case type
The white space surrounding lower case type makes more distinctive

shapes

Upper case type should be restricted to short headlines

Typography guidelines

THE POWER OF TYPE The Power of Type

Best Practices in Data Visualization – ComSciConCan Jenkinson

Enhance visual distinctions

Black Black Black Bold Black Medium Black Light Bold Bold Bold Medium Medium Light Bold Light

Best Practices in Data Visualization – ComSciConCan Jenkinson

Guidelines

Visual attribute Non-data Data

Line thickness thin thick Size small big Luminance contrast decrease increase Colour saturation decrease increase Enclosure no yes

SLIDE 23

Best Practices in Data Visualization – ComSciComCan Jenkinson

Encoding: choose visual representation that matches and supports data
Context: provide adequate context to make meaningful comparisons
Clarity: reduce non-data ink & highlight data ink

In sum…

rinse & repeat…. design is an iterative process

Alberto Cairo Edward Tufte Points of View & Points of Significance Colin Ware Cole Nussbaumer Knaflic

Data Visualization Toolkits

Best Practices in Data Visualization – ComSciComCan Jenkinson

Data rarely comes in a ready-to-visualize form
Approaches to cleaning data:
Writing custom scripts
Manual manipulation (Excel)
Data Wrangler (Trifacta): https://www.trifacta.com/
Open Refine: http://openrefine.org
Tableau Prep: https://www.tableau.com/products/prep

Data cleaning

SLIDE 24

Best Practices in Data Visualization – ComSciComCan Jenkinson

Tableau: https://www.tableau.com
D3.js: https://d3js.org
Circos: http:/

/circos.ca

Plotly: https://plot.ly/
Datawrapper: https://www.datawrapper.de
RawGraphs: https://rawgraphs.io

Data visualization

Best Practices in Data Visualization – ComSciConCan Jenkinson

Data https:/ /uoft.me/CSCdata

Best Practices in Data Visualization – ComSciConCan Jenkinson

Visualizations

Bladder Brain/CNS Breast Cervix Colorectal Esophagus Hodgkin lymphoma Kidney and renal pelvis Larynx Leukemia Liver Lung and bronchus Melanoma Multiple myeloma Non-Hodgkin lymphoma Oral Ovary Pancreas Prostate Stomach Testis Thyroid Uterus (body, NOS)

Datawrapper RawGraphs Plotly

SLIDE 25

Best Practices in Data Visualization – ComSciConCan Jenkinson

Providing insight

I II III IV x y x y x y x y 10 8.04 10 9.14 10 7.46 8 6.58 8 6.95 8 8.14 8 6.77 8 5.76 13 7.58 13 8.74 13 12.74 8 7.71 9 8.81 9 8.77 9 7.11 8 8.84 11 8.33 11 9.26 11 7.81 8 8.47 14 9.96 14 8.1 14 8.84 8 7.04 6 7.24 6 6.13 6 6.08 8 5.25 4 4.26 4 3.1 4 5.39 19 12.5 12 10.84 12 9.13 12 8.15 8 5.56 7 4.82 7 7.26 7 6.42 8 7.91 5 5.68 5 4.74 5 5.73 8 6.89

Anscombe’s quartet

Best Practices in Data Visualization – ComSciComCan Jenkinson

Same
Mean (x and y)
Variance
Correlation
Regression

Why visualize?

Anscombe’s quartet

I II III IV x y x y x y x y 10 8.04 10 9.14 10 7.46 8 6.58 8 6.95 8 8.14 8 6.77 8 5.76 13 7.58 13 8.74 13 12.74 8 7.71 9 8.81 9 8.77 9 7.11 8 8.84 11 8.33 11 9.26 11 7.81 8 8.47 14 9.96 14 8.1 14 8.84 8 7.04 6 7.24 6 6.13 6 6.08 8 5.25 4 4.26 4 3.1 4 5.39 19 12.5 12 10.84 12 9.13 12 8.15 8 5.56 7 4.82 7 7.26 7 6.42 8 7.91 5 5.68 5 4.74 5 5.73 8 6.89

SLIDE 26

Best Practices in Data Visualization – ComSciConCan Jenkinson

Why visualize?

I II III IV x y x y x y x y 10 8.04 10 9.14 10 7.46 8 6.58 8 6.95 8 8.14 8 6.77 8 5.76 13 7.58 13 8.74 13 12.74 8 7.71 9 8.81 9 8.77 9 7.11 8 8.84 11 8.33 11 9.26 11 7.81 8 8.47 14 9.96 14 8.1 14 8.84 8 7.04 6 7.24 6 6.13 6 6.08 8 5.25 4 4.26 4 3.1 4 5.39 19 12.5 12 10.84 12 9.13 12 8.15 8 5.56 7 4.82 7 7.26 7 6.42 8 7.91 5 5.68 5 4.74 5 5.73 8 6.89

Anscombe’s quartet

4 6 9 11 13 4 8 12 16 20 4 6 9 11 13 4 8 12 16 20 4 6 9 11 13 4 8 12 16 20 4 6 9 11 13 4 8 12 16 20 Francis J. Anscombe, Graphs in Statistical Analysis. The American Statistician, vol. 27, no. 1, pp. 17–21, 1973 10 8.04 6.95 13 7.58 8.81 8.33 14 9.96 7.24 4.26 12 10.84 4.82 5.68

“…If a picture is only worth a thousand words, we're screwed.”

Eric Lander Professor of Biology, MIT VIZBI 2011, Opening Remarks

Best Practices in Data Visualization – ComSciComCan Jenkinson

Visual mapping (encoding data)
Providing adequate context
Balancing clarity & aesthetics

Data Visualization in a nutshell Encoding Data

SLIDE 27

Best Practices in Data Visualization – ComSciComCan Jenkinson

Visual representation of data should be consistent with the numerical

representation

Title Text

Best Practices in Data Visualization – ComSciComCan Jenkinson

Nominal
name, type, category
eg. mammals, reptiles, birds
Ordinal
integer sequence
eg. first, second, third
happy, very happy, ecstatic

Data types

Best Practices in Data Visualization – ComSciComCan Jenkinson

Interval
gap in values
eg. every three months
Ratio
real numbers; zero as reference
45.7 out of 100

Data types

Best Practices in Data Visualization – ComSciComCan Jenkinson

Spatial
eg. maps, GIS, directions
scalar fields
Narrative
eg. assembly sequence, process

Data features

SLIDE 28

Best Practices in Data Visualization – ComSciConCan Jenkinson

Data features

1d 2d 3d

Best Practices in Data Visualization – ComSciComCan Jenkinson

The way in which data is mapped to visual structures
Every visualization can be described as a set of mappings:
from data items to visual marks
from data attributes to visual channels

Visual encoding

Best Practices in Data Visualization – ComSciComCan Jenkinson

Data Items:
Data Marks: the basic visual units that represent data objects visually
Data Attributes:
Visual Channels: the visual variables we can use to represent

characteristics of these objects

Title Text

Best Practices in Data Visualization – ComSciConCan Jenkinson

Marks

From Enrico Bertini

SLIDE 29

Best Practices in Data Visualization – ComSciConCan Jenkinson

Channels

From Enrico Bertini Best Practices in Data Visualization – ComSciConCan Jenkinson

Cleveland & McGill’s Perceptual Task Scale

Allows more accurate judgments Allows more generic judgments Position along a common scale Position along nonaligned scales Length Colour saturation Shading Curvature Volume Area Angle Direction

From The Functional Art, Alberto Cairo Best Practices in Data Visualization – ComSciComCan Jenkinson

Tasks are grouped according to how well you can perceive differences in

the data

Cleveland and McGill’s Perceptual Task Scale

Best Practices in Data Visualization – ComSciComCan Jenkinson

People are not good at making visual angular distinctions
Pie charts are sometimes rolled out to encode 1 or 2 numbers; usually a

very low data density!

A basic example: the pie chart

SLIDE 30

Best Practices in Data Visualization – ComSciConCan Jenkinson

Pie charts

2002 2003 2004 2005 2006 2007

Best Practices in Data Visualization – ComSciConCan Jenkinson

Pie charts

7% 8% 10% 11% 29% 35%

2002 2003 2004 2005 2006 2007

“There are three kinds of lies: lies, damned lies, and statistics”

Benjamin Disraeli

Best Practices in Data Visualization – ComSciComCan Jenkinson

Misalignments between graphic elements and the data they are

intended to represent

Lies, damned lies…

SLIDE 31

Best Practices in Data Visualization – ComSciConCan Jenkinson

Misleading use of area

Best Practices in Data Visualization – ComSciConCan Jenkinson

Misleading use of area

Best Practices in Data Visualization – ComSciConCan Jenkinson

Misleading use of area

Best Practices in Data Visualization – ComSciConCan Jenkinson

Misleading use of area…

SLIDE 32

Best Practices in Data Visualization – ComSciComCan Jenkinson

A common mistake for chart design is to scale an area by two sides at

the same time, producing a quadratic effect for a linear change

Linear vs. quadratic change

Tie area of the white square = a

2

a = 80 Tie area of the white square =6,400 px To double the area of white square =12,800 px a = 160 Tie area of the greysquare a = 25,600 px

2

...four times that of the white square Tie area of red square is twice that of white square √12,800 = 113 a = 113

Best Practices in Data Visualization – ComSciConCan Jenkinson

Linear vs quadratic change

The area of the blue circle is equal to πr2 (20,106) The area of the red circle is equal to πr2 (80, 424)

r = 80 r = 160

Best Practices in Data Visualization – ComSciConCan Jenkinson

Size encoding

Height Area Volume

Best Practices in Data Visualization – ComSciComCan Jenkinson

Modifications to the X or Y axis in an attempt to make differences or

change appear to be more dramatic

Data represented out of context does not allow for adequate comparison

How statistics lie

SLIDE 33

Best Practices in Data Visualization – ComSciConCan Jenkinson

The disappearing baseline

Best Practices in Data Visualization – ComSciConCan Jenkinson

The disappearing baseline

Best Practices in Data Visualization – ComSciConCan Jenkinson

Reversing the x-axis

SLIDE 34

Best Practices in Data Visualization – ComSciConCan Jenkinson

Reversing the x-axis

Best Practices in Data Visualization – ComSciConCan Jenkinson

No defined y-axis

Best Practices in Data Visualization – ComSciConCan Jenkinson

No defined y-axis Context

SLIDE 35

Best Practices in Data Visualization – ComSciComCan Jenkinson

Without context we are unable to see the big picture
Without context we are unable to make meaningful comparisons

Context

Source: New York Times Source: New York Times Source: New York Times

SLIDE 36

Best Practices in Data Visualization – ComSciConCan Jenkinson

Data in and out of context

Best Practices in Data Visualization – ComSciConCan Jenkinson

Data in and out of context

Best Practices in Data Visualization – ComSciConCan Jenkinson

Data in and out of context

Best Practices in Data Visualization – ComSciConCan Jenkinson

Historical context

SLIDE 37

Best Practices in Data Visualization – ComSciConCan Jenkinson

Historical context Clarity & Aesthetics

Best Practices in Data Visualization – ComSciComCan Jenkinson

The important of the information should match the salience of the

channel

Effectiveness

Kim OY, et al. (2012) Higher levels of serum triglyceride and dietary carbohydrate intake are associated with smaller LDL particle size in healthy Korean women. Nutrition Research and Practice 6:120-125

SLIDE 38

Cawley S, et al. (2004) Unbiased mapping of transcription factor binding sites along human chromosomes 21 and 22 points to widespread regulation of noncoding RNAs. Cell 116:499-509

French Polynesia Chad Afghanistan Congo (Kinshasa) Burundi Cambodia Uganda Mali Ethiopia Malawi Burkina Faso Central African Republic Somalia Rwanda Laos Niger Nepal Lesotho Tanzania Madagascar Guinea Eritrea Comoros Liberia Gambia, The Sierra Leone Haiti Zambia Mozambique Burma(Myanmar) Guinea-Bissau Bangladesh Kenya Sudan Ghana Benin Cote d'Ivoire (IvoryCoast) Kiribati Solomon Islands Cameroon Togo Senegal Bhutan Vanuatu Sao Tome and Principe Paraguay Sri Lanka Cape Verde Turks and Caicos Islands Nigeria Pakistan Samoa Philippines Papua New Guinea Nicaragua

W. Sahara

Yemen Zimbabwe Guatemala Swaziland El Salvador Kyrgyzstan Mauritania Georgia Honduras Morocco Peru Tajikistan Vietnam U.S. Pacific Islands India Indonesia Albania Namibia Tonga Dominica Bolivia Costa Rica Colombia Fiji Congo (Brazzaville) Saint Vincent/Grenadines Moldova Angola Saint Helena Uruguay Ecuador Dominican Republic Egypt Brazil Maldives Tunisia Saint Lucia Botswana Cuba Syria Algeria Mongolia Grenada Gabon Mauritius Cook Islands Turkey North Korea Jordan Belize Armenia Macedonia Saint Kitt and Nevis Thailand Latvia Bosnia and Herzegovina Chile Mexico Argentina Djibouti Virgin Islands British Suriname Jamaica Lithuania Romania Uzbekistan Panama China Croatia Former Serbia and Montenegro Azerbaijan Macau Barbados Portugal Hungary Venezuela Switzer- land Sweden France Bulgaria Belarus Malaysia Slovakia Ukraine Iran Montserrat Malta Poland Italy Equatorial Guinea Slovenia Libya Cayman Islands Spain Austria New Zealand Aruba Antigua and Barbuda United Kingdom Japan Norway Israel Greece Turkmenistan South Africa Bermuda Germany Korea, South Puerto Rico American Samoa Denmark Greenland Guam Finland Oman Seychelles Czech Republic Cyprus Iceland Ireland New Caledonia Saint Pierre and Miquelon Russia Hong Kong Taiwan Nauru Kazakhstan Estonia Belgium Faroe Islands Saudi Arabia Netherlands The Bahamas Canada United States Australia Luxembourg Brunei Kuwait

Singapore

United Arab Emirates Bahrain Trinidad and Tobago

Netherlands Antilles Virgin Islands, U.S. Gibraltar

Iraq Lebanon

Qatar

Serbia Nepal Chad Wake Island U.S. Pacific Islands Timor-Leste (East Timor) Solomon Islands Papua New Guinea Maldives Laos French Polynesia Cook Islands Brunei Bhutan New Zealand American Samoa Afghanistan Cote d’Ivoire Western Sahara Tanzania Seychelles Sao Tome and Principe Saint Helena Guinea-Bissau Madagascar Mozambique Equatorial Guinea Congo (Brazzaville) Comoros Central African Republic Cape Verde Burkina Faso Yemen Palestine Oman Lebanon Bahrain Tajikistan Moldova Kyrgyzstan Estonia Azerbaijan Armenia Macedonia Iceland Hungary Gibraltar Former Serbia and Montenegro Faroe Islands Cyprus Bosnia and Herzegovina Albania Dominica Virgin Islands, U.S. Turks and Caicos Islands Trinidad and Tobago Suriname Saint Vincent/Grenadines Saint Lucia Saint Kitts and Nevis Guatemala Nicaragua Netherlands Antilles Martinique Guadeloupe French Guiana Falkland Islands (Islas Malvinas) El Salvador Jamaica Costa Rica Cayman Islands Sierra Leone Paraguay Bahamas, The Saint Pierre and Miquelon Antigua and Barbuda Greenland Bermuda New Caledonia Virgin Islands, British Tonga Timor-Leste (East Timor) Tonga Samoa Nauru Uganda Tunisia Somalia Senegal Rwanda Niger Liberia Lesotho Guinea Gabon Eritrea Burundi Malta Haiti Panama Grenada Guyana Belize Honduras Aruba Zambia Togo Swaziland Namibia Morocco Mauritius Mauritania Mali Malawi Kenya Ghana Gambia Ethiopia Djibouti Cameroon Botswana Benin Angola Macau Cambodia Zimbabwe Turkmenistan Mongolia Luxembourg North Korea Vietnam Thailand

Taiwan

Singapore Philippines Pakistan Malaysia

South Korea Japan

Indonesia

India

Hong Kong

China

Sri Lanka Burma Bangladesh Fiji Guam

Australia

Reunion Sudan

South Africa Egypt

Congo (Kinshasa) Algeria United Arab Emirates Syria Saudi Arabia Qatar Libya Nigeria Kuwait Jordan Israel Iraq Iran Uzbekistan

Ukraine Russia

Lithuania Kazakhstan Georgia Belarus United Kingdom

Turkey

Switzerland Sweden

Spain

Slovakia Romania Portugal Poland Norway Netherlands Slovenia

Italy

Greece

Germany France

Finland Denmark Czech Republic Croatia Bulgaria Belgium Austria Venezuela Chile Peru Ecuador Colombia Brazil Bolivia Montserrat Uruguay Argentina Cuba Barbados Puerto Rico

United States

Dominican Republic

Mexico

Latvia Ireland

Canada

EUROPE

Total Carbon Emissions by Nation

Tracking Carbon Emissions

KEY

Per Capita Carbon Emissions by Nation

ASIA MIDDLE EAST AFRICA NORTH AMERICA CENTRAL AMERICA SOUTH AMERICA OCEANIA CARIBBEAN NOTE: BASED ON 2007 DATA. SOURCES: U.S. ENERGY INFORMATION ADMINISTRATION DESIGN: STANFORD KAY STUDIO.COM Chinaí s total emissions lead the world, but when diluted by its huge population, its ranking drops down the per capita list. Guyana The United States is

no. 2 for total emissions

but Americans shrink down to a respectable rank in line with other industrialized citizens. Tiny Gibraltar tops the per capita list due to its need to import most manufactured goodsó a reality also seen in many small island nations.

A footprint comparison of total carbon dioxide emissions by nation and per capita shows thereí s plenty of room for smaller countries to reduce their carbon footprints. By Stanford Kay

°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°°

“The data-ink ratio is the proportion of ink that is used to present actual data compared to the to the total amount of ink used in the entire display.”

Edward Tufte

SLIDE 39

Data/Ink = ink used for data total ink used in graphic

“The data-ink ratio is the proportion of ink that is used to present actual data compared to the to the total amount of ink used in the entire display.”

Best Practices in Data Visualization – ComSciComCan Jenkinson

Data arranged in columns and rows
Data encoded as words and numbers

Anatomy of a table

Time Topic Lecturer 2:10 pm The fine art of torture Ramsay Bolton 3:00 pm Break

3:10pm

Lessons in servitude Theon Greyjoy

Best Practices in Data Visualization – ComSciComCan Jenkinson

When “looking up” is important
When precision and value comparison are required
When units vary

When to use tables

Mortgage Rate Bank Type 5.2% EDQ Fixed 5% BMA Variable 4.8% INF Variable

Best Practices in Data Visualization – ComSciComCan Jenkinson

Are the grids necessary?

Design guidelines

Mortgage Rate Bank Type 5.2% EDQ Fixed 5% BMA Variable 4.8% INF Variable

SLIDE 40

Best Practices in Data Visualization – ComSciComCan Jenkinson

Usually, no

Design guidelines

Mortgage Rate Bank Type 5.2% EDQ Fixed 5% BMA Variable 4.8% INF Variable

Best Practices in Data Visualization – ComSciComCan Jenkinson

Wide tables sometimes need a graphic device to aid horizontal scanning

Design guidelines

Train Destination Time 23 Poughkeepsie 1:00 2:00 3:00 4:00 5:00 6:00 7:00 48 Timbuctoo 1:15 2:15 3:15 4:15 5:15 6:15 7:15 15 Wawa 1:30 2:30 3:30 4:30 5:30 6:30 7:30

Best Practices in Data Visualization – ComSciComCan Jenkinson

Wide tables sometimes need a graphic device to aid horizontal scanning

Design guidelines

Train Destination Time 23 Poughkeepsie 1:00 2:00 3:00 4:00 5:00 6:00 7:00 48 Timbuctoo 1:15 2:15 3:15 4:15 5:15 6:15 7:15 15 Wawa 1:30 2:30 3:30 4:30 5:30 6:30 7:30

Best Practices in Data Visualization – ComSciComCan Jenkinson

Wide tables sometimes need a graphic device to aid horizontal scanning

Design guidelines

Train Destination Time 23 Poughkeepsie 1:00 2:00 3:00 4:00 5:00 6:00 7:00 48 Timbuctoo 1:15 2:15 3:15 4:15 5:15 6:15 7:15 15 Wawa 1:30 2:30 3:30 4:30 5:30 6:30 7:30

SLIDE 41

Best Practices in Data Visualization – ComSciConCan Jenkinson

Milton GO Bus Service

Toronto Union->Mississauga->Meadowvale-> Milton zn Table Monday to Friday (excluding holidays) 2 Toronto-Union Bus Term 07 20 08 20 09 20 10 20 10 50 11 20 12 00 12 10 12 20 2 Union Station ! ! ! ! ! ! ! ! ! 3 Kipling GO Station ! ! ! ! ! ! ! ! ! 11 Dixie GO Station 07 40 08 40 09 40 10 40 11 10 11 40 12 20 ! ! 11 Cooksville GO Station 07 50 08 50 09 50 10 50 11 20 11 50 12 30 ! ! 20 Square One-Rathbum@D.O.Y. 08 00 09 00 10 00 11 00 11 30 12 00 12 40 ! ! 12 Erindale GO Station 08 05 09 05 10 05 11 05 11 35 12 05 12 40 ! 21 Streetsville GO Station 09 15 10 15 11 15 12 15 12 50 !

Best Practices in Data Visualization – ComSciConCan Jenkinson

Milton GO Bus Service

Toronto Union ! Mississauga ! Meadowvale ! Milton

Zone & Station Monday to Friday (excluding holidays) 02 Toronto-Union Bus Term 07:20 08:20 09:20 10:20 10:50 11:20 12:00 12:10 12:20 02 Union Station ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ 03 Kipling GO Station ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ 11 Dixie GO Station 07:40 08:40 09:40 10:40 11:10 11:40 12:20 ¯ ¯ 11 Cooksville GO Station 07:50 08:50 09:50 10:50 11:20 11:50 12:30 ¯ ¯ 20 Square One-Rathburn@D.O.Y. 08:00 09:00 10:00 11:00 11:30 12:00 12:40 ¯ ¯ 12 Erindale GO Station 08:05 09:05 10:05 11:05 11:35 12:05 12:40 ¯ 21 Streetsville GO Station 09:15 10:15 11:15 12:15 12:50 ¯

Best Practices in Data Visualization – ComSciComCan Jenkinson

Our ability to perceive colour is surprisingly limited
Much of the information we perceive is derived from luminance (light and

shadow)

Colour guidelines

SLIDE 42

Best Practices in Data Visualization – ComSciComCan Jenkinson

To attract attention
To group elements
To indicate meaning
To enhance aesthetics

Fundamental uses of colour

Best Practices in Data Visualization – ComSciComCan Jenkinson

Reserve saturation for areas requiring attention
Background colours should be low in saturation
If you don’t need colour don’t use it

Colour recommendations

SLIDE 43

Best Practices in Data Visualization – ComSciComCan Jenkinson

Harmonizing text and line-drawing requires sensitive appraisals of

interaction effects

Unless necessary, avoid boxing in text as this activates negative space

between the word and the boxes

Typography guidelines

Best Practices in Data Visualization – ComSciComCan Jenkinson

UPPER CASE TYPE IS MORE DIFFICULT TO READ THAN lower case type
The white space surrounding lower case type makes more distinctive

shapes

Upper case type should be restricted to short headlines

Typography guidelines

THE POWER OF TYPE The Power of Type

Best Practices in Data Visualization – ComSciConCan Jenkinson

Enhance visual distinctions

Black Black Black Bold Black Medium Black Light Bold Bold Bold Medium Medium Light Bold Light

Best Practices in Data Visualization – ComSciConCan Jenkinson

Guidelines

Visual attribute Non-data Data

Line thickness thin thick Size small big Luminance contrast decrease increase Colour saturation decrease increase Enclosure no yes

SLIDE 44

Best Practices in Data Visualization – ComSciComCan Jenkinson

Encoding: choose visual representation that matches and supports data
Context: provide adequate context to make meaningful comparisons
Clarity: reduce non-data ink & highlight data ink

In sum…

rinse & repeat…. design is an iterative process

Alberto Cairo Edward Tufte Points of View & Points of Significance Colin Ware Cole Nussbaumer Knaflic

Data Visualization Toolkits

Best Practices in Data Visualization – ComSciComCan Jenkinson

Data rarely comes in a ready-to-visualize form
Approaches to cleaning data:
Writing custom scripts
Manual manipulation (Excel)
Data Wrangler (Trifacta): https://www.trifacta.com/
Open Refine: http://openrefine.org
Tableau Prep: https://www.tableau.com/products/prep

Data cleaning

SLIDE 45

Best Practices in Data Visualization – ComSciComCan Jenkinson

Tableau: https://www.tableau.com
D3.js: https://d3js.org
Circos: http:/

/circos.ca

Plotly: https://plot.ly/
Datawrapper: https://www.datawrapper.de
RawGraphs: https://rawgraphs.io

Data visualization

Best Practices in Data Visualization – ComSciConCan Jenkinson

Data https:/ /uoft.me/CSCdata

Best Practices in Data Visualization – ComSciConCan Jenkinson

Visualizations

Bladder Brain/CNS Breast Cervix Colorectal Esophagus Hodgkin lymphoma Kidney and renal pelvis Larynx Leukemia Liver Lung and bronchus Melanoma Multiple myeloma Non Hodgkin lymphoma Oral Ovary Pancreas Prostate Stomach Testis Thyroid Uterus (body, NOS)

Datawrapper RawGraphs Plotly