an introduction to

VISUALIZING DATA by joel laumans

 an introduction to VISUALIZING DATA  | iii

Table of Contents
AN INTRODUCTION TO VISUALIZING DATA by Joel Laumans

1 Introduction 1
Definition
Purpose

2 Data visualizations 2

3 Examples 3

4 The data 5
Examine the data

5 Data visualization patterns 8

6 Revealing the data 12

Preattentinve variables

7 Multimedia 16
Screen resolutions
User interaction

8 Tools and further reading 20

iv |  Joel Laumans   | 1

Foreword

The purpose of this document is to provide an

introduction to the theory behind visualizing
data. After studying the works of many talented
1 Introduction

people I decided to summarize the key points To create a truly powerful data visualization a
of information into this single paper. If you combination of artistic, statistical, and math-
found this document interesting please take ematical skills are required - this is most likely
some time to look at the list of resources that the reason why the first multivariate statistical
I used (see Chapter 8) because I could never graphics only appeared late in the 18th cen-
have created this without the excellent work tury. Over time the use of data visualization
done by others. has become continually more popular; partly
because the tools to create data graphics are
If you have any comments or feedback please
readily available, but also, because there is
feel free to contact me.
an urgency to communicate information both
quickly and effectively as possible.
Joel Laumans The purpose of this document is to provide an
introduction to data visualization by exploring
User Experience Designer
the purpose, the requirements, and methods of
[email protected] visualizing data.
http://www.piksels.com/
http://creatinginspiration.net/
http://www.linkedin.com/in/joellaumans
2 |  Joel Laumans an introduction to VISUALIZING DATA  | 3

2 Data visualizations
3 Examples

Name popularity in the United States of America

Definition Purpose
This is a collection of maps visualizing the popularity of names
in different states of the United States. The color (pink vs blue)
Data visualizations, also known as data graph- Creating a data visualization is more than sim- represents the gender of the name, and the saturation of the
ics, can be best explained by quoting Edward ply translating a table of data into a visualiza- color represents the popularity of the name in the respec-
Tufte: tion. Data visualizations should communicate tive region. This is a very simple and powerful visualization
data in the most effective way; to truly reveal because we can easily conclude that ‘Alex’ is popular through
the data they should be quick, accurate, and almost the entire country, while ‘Colin’ is more popular in the
“Data graphics visually display mea- powerful. Creating visuals can easily sum- North Eastern states.
marize and communicate data to other people
sured quantities by means of the - making even the largest or most complicated
combined use of points, lines, a co- sets of data understandable.
ordinate system, numbers, symbols,
words, shading, and color.” Usage
Campaign Finances

A common misconception is that data visual- Data visualizations can take many different
This is a screenshot of an interactive visualization made by The
izations are the same as information graphics forms depending on the information that is
New York Time’s to visualize the which regions were financing
(infographics). It is important to understand being communicated - from simple bar charts
the US presidential candidates. Each circle represents a major
that data visualizations always communicate that communicate rising oil prices to interac-
city, and the size of the circle represents the amount of money
a message by visualizing quantifiable data tive applications that analyze website visitor
donated. At the bottom is a bar chart which shows the amount
objectively, while infographics can be used to data.
of money (vertical) donated per week (horizontal). What makes
communicate any information at all (usually
it very powerful, was the element of interaction - users can
with a specific goal); regardless of whether it is
select which candidate they want to view, the time span, and
quantifiable or not.
retrieve extra data by clicking on the circles.
4 |  Joel Laumans an introduction to VISUALIZING DATA  | 5

Distribution of the world’s water

4
This a great example of how visualizing data can be used to

The data
communicate a very powerful message. Read from top to bot-
tom, this visual tells us that only 2,5% of the world’s water is
fresh water, and of those 2,5% only a small portion is actually
available for human use because the majority is frozen in gla-
ciers; a fact that most people are unaware of, but communi-
cated clearly in this visualization..

There are countless methods of how to cre-

ate data visualizations - they will vary greatly
depending on the content of the data, as well
as the purpose of the visualization. Whichever
form it takes, the most important is to maintain
graphical excellence, explained by Tufte as:

“Excellence in statistical graphics

consists of complex ideas commu-
nicated with clarity, precision, and
efficiency.”

According to Tufte’s checklist, all data visual-

izations should:
♦♦ Show the data
♦♦ Be accurate (don’t distort the data)
♦♦ Make large data sets coherent
Virtual Water
♦♦ Serve a clear purpose
This is another visualization created to raise awareness about
♦♦ Reveal the data at different levels (overview
excessive water usage. It visualizes the amount of water that
versus detailed)
is used in the production of certain products. The blue water
drops represent the amount of water used per product: for ♦♦ Encourage the viewer to compare different
example it takes 1000 liters of water to produce 1 liter of milk, pieces of data
or 4500 liters of water for a single piece of steak. Very simple
and effective.
6 |  Joel Laumans an introduction to VISUALIZING DATA  | 7

Examine the data

Every set of data can be visualized in multiple Examples of how data can be organized Ben Fry’s “Seven Stages of Visualizing Data”
ways, some more effective than others. In
order to create an effective data visualization Visualizing complex data sets often requires insights from diverse fields of knowledge, such
as statistics, data mining, graphic design, et cetera. Ben Fry suggests a seven stage design
the data first that to be understood. Therefore,
process, reconciling all stages into a single process.
the first step in creating a data visualization is
to examine the data. Basic questions that need
1. Aquire
to be answered are:
Obtain the data, whether from an Excel document, an XML feed, et cetera.
♦♦ What is the data?
♦♦ What are the relationships between the 2. Parse
variables? linear tabular Data will not always be organized ideally for visualizing it. Give your data structure by order-
ing it into categories.
♦♦ How is the data organized?
♦♦ What needs to be communicated?
3. Filter
Be careful to prevent information overload, remove all but the data of interest
Before thinking about how the visualization
should look, it is important to have answered 4. Mine
these questions because they will determine Apply methods from statistics or data mining as a way to find patterns or meaning in the
the form of the data visualization. data.
One of the most important parts is understand- hierarchical networked
ing how the data is organized and related to 5. Represent
each other. For example, if you want to com- Choose a basic visual model to visual the data. (see Chapter 5)
municate company growth over time a simple
line chart might be the most effective - but to
6. Refine
communicate population density around the
world it might be more effective to use a carto- Improve the basic representation to make it more clear and more visually engaging. (see
Chapter 6)
graphic (map) visual.
There are countless possibilities of how to
7. Interact
visualize data; but there are many design
patterns which are proven to be effective for geographic other
Add methods for manipulating the data. Allows users to control what they see or even pos-
sibly how they see it.
specific types of data. The following section
provides examples of data visualization pat-
terns and a short rationale for when to use
them.
8 |  Joel Laumans an introduction to VISUALIZING DATA  | 9

Pie charts

5
Proportions

Simple pie charts are the most common visual

Data visualization patterns

Proportions are used when
used to compare proporational data. They
the data represents parts of
give viewers a very quick understanding of the
a whole.
distribution of the data. Pie charts are not use-
ful when comparing many pieces of data with
relatively close values.

Independent quantities Bar charts Ring charts

Simple bar charts are the most common form Similar to pie charts; ring charts are used to
Comparing the values of
of data visualizations. Typically they only dis- visualize the distribution of a data set. The
independent variables
play different quantities of single-variable data. advantage is they compare similar data sets.
However other variations, such as stacked bar The alternative would be to place multiple
charts or multi-set bar charts can be used to pie charts next to each other, this can also be
compare multiple variables using bars. viewed as a space-saver.

Correlations
Continuous quantities Line graphs Scatterplots
Line graphs are created by plotting points on a These are used when each They are created by plotting independent
For data that is continuous,
Cartesian grid, usually with the horizontal axis piece of data has two quanti- points on a Cartesian grid. Scatter plots are of-
for example when visualizing
representing time. They are very powerful be- fiable variables which can be ten used to find the relationship between data
data over a period of time.
cause without looking at the specific data, they plotted on a grid. or to reveal information such as trends within
show how a variable develops over time (from the data which are not easily visible when in a
left to right). table. Only works with two dimensional data.

Stacked area charts Bubble charts

Similar to scatterplots; however, bubble charts
Similar to line charts, however with the added
display more dimensions of data by varying
value of filled areas. The data that is stacked
size (or also color, texture, etc) of the bubbles.
adds up to a total of all variables combined. For
It therefore can display multiple dimensions of
example a business might use stacked area
data in a two dimensional display.
charts to visualize their total income, with each
stacked area a different income channel.
10 |  Joel Laumans an introduction to VISUALIZING DATA  | 11

Hierarchies Tree diagrams then DVD sales. Each smaller arrow would be
representing a source of income.
Tree diagrams are often used when wanting to
Used when the data has a
strict hierarchy that needs to
represent the strict hierarchy of data. They are Data visualization patterns (cont.)
most often used to represent strict hierarchies
be communicated. The aforementioned examples represent only
such as family trees or how data is stored in a
a small number of the most popular types of
computer system.
visualizations. To find more information about
the different types of data visualizations refer
to the references listed at the end of this docu-
ment.
There are many advantages to using existing
Networks
Diagram maps patterns for data visualization. They have been
proven to work effectively for a specific type of
Network visualizations are These visualizations are used to primarily
dataset. Furthermore, people are familiar with
used when the most impor- represent the connections between differ-
reading these types of visualizations, making
tant feature of the visualiza- ent nodes or points. Their purpose is to show
them easier to understand.
tion is to show which data is which points are connected to each other.
connected to each other, as Common examples of diagram maps are metro Unfortunately, many data sets have unique
opposed to how. maps and social network visualizations. characteristics which will force you to come up
with new ways of communicating the data in
the most effective and powerful way possible.
When one single data visualization pattern will
Cartographic not be adequate, the solution many times is to
Maps create a combination of several data patterns
For data that is relevant to to create a single message. Each pattern can
Maps are used when the data is related to a
specific locations or regions then be used to represent specific variables in
specific location (for example a city, or coun-
which can be plotted on a the data.
try). The advantage is that their spatial repre-
map.
sentation directly relates to a real-world situa- The two visualizations on the right are good
tion. However at times can be difficult to read. examples of how to combine visualization pat-
terns. The top visualization is a combination of
a cartographic, bubble chart, and network dia-
gram. The bottom is a combination of bubble
Flows chart and pie charts plotted onto a matrix.

When the data is part of a

Sankey diagrams
process, it can be visualized Sankey diagrams are composed of several
using flow diagrams. smaller ‘arrows’ or channels, which merge
together into one large channel. For example,
a sankey could be used to visualize a movie’s
revenue, first movie tickets, then merchandise,
12 |  Joel Laumans an introduction to VISUALIZING DATA  | 13

Now take a look at the values in the table and

6
find all numbers greater than 1.0.

Revealing the data

Creating a powerful data visualization is not

about simply translating a table of data into
Preattentinve variables
Table 1: Find the values larger than 1.0
a visual graphic, it is about communicating There are certain visual features in design that
the meaning of the data. Choosing the most will work preattentively; they will communicate
adequate visualization design pattern is an something about the design before the user There is no visual aid to help us find the values
important step because it will immediately pays conscious attention to it. Jennifer Tidwell greater than 1.0, forcing us to read all of the
tell users about how the data is organized and has coined these as preattentive variables. data and understand it. Now look at the follow-
what you are trying to communicate about the ing table.
data. However, it is not only the type of visual- A powerful data visualization should work
ization pattern chosen - but also the design of quickly and effectively; therefore the visual
the individual elements that play an important design should help reveal the data to the users.
role in communicating information to others.
In Tidwell’s book, Designing Interfaces, she
provides the following examples.
Take a look at the following group of dots and
find all of the blue dots. Table 2: Find the values larger than 1.0

Finding the values has become much easier by

simply changing some of the visual features we
can significantly improve the search times for
users.
On the following pages all eight of Tidwell’s
preattentive variables will be demonstrated.

Quick, right? Even if we increased the number

of red dots, finding the blue dots will be just as
quick, because color works preattentively!
14 |  Joel Laumans an introduction to VISUALIZING DATA  | 15

Color hue Color brightness Color saturation Texture

Position Orientation Size Shape

In essence, all display patterns use any single

or combination of these variables to visualize
data. Properly applying these visual variables
is what allows large sets of data to be quickly
and accurately understood in data visualiza-
tions.
16 |  Joel Laumans an introduction to VISUALIZING DATA  | 17

7 Multimedia

Until recently, all data visualizations were

static and predefined; however, many modern
Screen resolutions
data visualizations are created using multime-
dia interfaces that expand the possibilities of
data visualizations. The most common screens where users will
come across data visualizations are desktop
When working with multimedia interfaces (as
monitors and mobile phones.
opposed to print), there are new factors which
can influence the effectiveness of communi- Modern mobile phones resolutions vary be-
cating information to users. The most limiting tween 144 x 176 pixels up until 320 x 480 pixels,
factor of multimedia is the relatively small with the most popular being 240 x 320 pixels.
amount of information that can be displayed on Only simple data visualizations can be dis- Mobile resolutions Desktop monitor resolutions

a screen, but the main advantage of using mul- played at this resolution. Comparing 20 differ-
timedia is that visualizations can be dynamic, ent pieces of data becomes difficult.
animated, and allow for user interaction.
Typical desktop computer resolutions are increase the effectiveness of the data visual- visualization, they can control and manipulate
much higher, ranging between 1024 x 768 pix- ization. When working with multimedia data how and what is displayed. There are many
els and 1600 x 1200 pixels, allowing for richer visualizations, the most effective resolutions forms of user interaction that can be applied
visualizations that communicate sets of data will be those of standard desktop monitors. to data visualizations. These forms are in two
that are larger and more complex. categories:
The graph on the following page illustrates the ♦♦ Data selection and filtering – The user can
effectiveness of a data visualization versus the control which data is displayed
screen resolution of the medium. Low resolu-
tions are a true limiting factor for displaying User interaction ♦♦ Data arrangement and navigation – The user
can control how the data is displayed or
data. The effectiveness of a data visualization
viewed.
increases quickly as the resolution increases,
however plateaus once reaching the upper Traditionally, data visualization was about Data selection and filtering will help users
limit of desktop monitor resolutions. The choosing the correct design pattern to visu- control precisely which data is being visual-
reason for the plateau is simply because, at alize the respective data to create the most ized. This will help users find data only relevant
this point, more information can be displayed meaning; however when we allow the user to what they are looking for, and help prevent
than a person can see at any single point in to interact with the data, it is not only about information overload.
time. This means that by being able to display how the data is displayed, but also about how
Data arrangement and navigation can help
more information than a user can see will not it behaves. When users interact with the data
18 |  Joel Laumans an introduction to VISUALIZING DATA  | 19

users find new meaning in the data. Simply dis- Categories of data visualizations
playing the same data in a different fashion can
help people come to new conclusions and see Static
different relationships between the data.
The data visualization does
“Each set of data has particular not change over time, and
display needs, and the purpose for offers no user interaction.
Data visualization becomes

which you’re using the data set has

a research instrument

just as much of an effect on those

Animated
needs as the data itself.” (Ben Fry)
The data visualization
Allowing users to control these two variables changes over time - using
instantly makes data visualizations more time as another dimension
powerful, simply because it makes them more or variable.
particular to a single user.
When using screens as a medium for display-
ing data visualizations, they can be organized
into four different categories: static, animated, Interactive Static Animated Interactive Direct manipulation

interactive, and direct manipulation

The user can interact with
The most powerful type are those that support
the visualization by altering
direct manipulation because they will let users
what data is viewed and / or discover new meaning in the data. By adding
immediately see how the variables they adjust
how it is viewed - however user interaction, the data visualizations have
influence the data that is being displayed.
the results of interactions been transformed from a static display of
These visualizations that which support direct
are not immediately visible. quantitative data into a tool for discovering new
manipulation, have the advantage that:
meaning and relationships in the data.
♦♦ Users can quickly learn the relationship
between different variables.
Direct manipulation
♦♦ Users can immediately see if their actions
are furthering their goals, and if not, they The user can interact with
can simply change the direction of their the visualization by altering
activity. what data is viewed and / or
how it is viewed - and the
In conclusion, when you take a traditional data
result of all interactions are
visualization and add the combination of user
immediately visible.
interaction and direct manipulation, it creates
a powerful formula for visualizing data. Giving
the user control over the visualization makes it
more meaningful to the user, and helps them
20 |  Joel Laumans an introduction to VISUALIZING DATA  | 21

Many Eyes

8
Tools

http://manyeyes.alphaworks.ibm.com/manyeyes/
Although simple data visual-

Tools and further reading ziations can be made using

software such as Microsoft
A research group at IBM created Many Eyes,
an online platform for data visualizations. Us-
Excel, Here are tools you can ing existing design patterns and data you can
use to create more complex experiment and create your own data visualiza-
data visualizations. tions. This is a great place to start for people
working with data visualizations for the first
Books The Visual Display of Quantitative time.
Information
Here is a short list of books
by Edward Tufte
for anyone interested in Processing
learning more about visual- This is the classic book about data graphics.
http://processing.org/
izing data. Tufte explains the history and theory behind
data graphics using more than 250 great ex- Processing is one of the most popular plat-
amples. This is a must have book for all people forms for visualizing data. This this the must
interested in visualizing data. have tool for people who are interested in
creating custom data visualizations with their
own data. The initial learning curve is quite
Designing Interfaes high because basic programming knowledge is
required. The best way to get started is to buy
by Jenifer Tidwell
the Ben Fry’s book Visualizing Data which pro-
This is a brilliant book on User Interface Design vides tutorials for people interested in learning
with handy information for people interested in Processing.
data visualizations. Tidwell has a whole chap-
ter dedicated to visualizing complex data in
Websites
which she explains how to organize and display Information aesthetics
data effectively.
These are two websites http://infosthetics.com/
that are frequently updated
with great examples of data
Visualizing Data visualizations and infograph-
by Ben Fry ics. They showcase the best
examples of data visualiza-
Visual Complexity
For those interested in creating complex data
tion out there.. http://www.visualcomplexity.com/vc/
visualizations, this is book is for you. This book
explains how to create data visualizations us-
ing Processing (an opensource platform for
data visualization). Each chapter is a tutorial
teaching you about data visualization as well as
how to use Processing.
22 |  Joel Laumans an introduction to VISUALIZING DATA  | 23

Credits and thanks

A large portion of this booklet is based on work

done by Edward Tufte, Ben Fry, and Jennifer
Tidwell; three people which have done remark-
able work for data visualization. Thank you.
If you find this booklet interesting, please con-
sider looking into the work of these people.
I would also like to thank the following people
for reviewing previous versions of this booklet
and giving me helpful feedback, I appreciate it.

Nathan Verril
User Experience Consultant
http://www.linkedin.com/pub/1/150/135

Bas Leurs
Lecturer Interaction Design at the Rotterdam
University of Applied Siences
http://www.linkedin.com/pub/0/330/308

Peter van Waart

Lecturer and Researcher at the Rotterdam
University of Applied Siences
http://www.linkedin.com/in/petervanwaart