DATA VISUALISATION

A practical guide to producing effective visualisations for

research communication

Rebecca Wolfe, 2014

Research Uptake Manager, RESYST Consortium

London School of Hygiene & Tropical Medicine
!
!

Email: [email protected]
Twitter: @RESYSTresearch
DATA VISUALISATION
A practical guide to producing effective visualisations for research communication

CONTENTS

1. INTRODUCTION TO DATA VISUALIATION

1.1 What is data visualisation?
1.2 The value of data visualisations

2. PLANNING A VISUALISATION
2.1 Start with the objective and audience
2.2 Explore and select the data
2.2 Choose the right type of chart

3. DESIGN TIPS FOR YOUR VISUALISATION

3.1 Represent the data accurately
3.2 Style the chart simply
3.3 Use colour with caution
3.4 Choose a clear font
3.5 Structure the story

4. TOOLS TO CREATE VISUALISATIONS

4.1 Online data visualisation tools
4.2 Blogs and resources
4.3 Books, articles and references
1. Introduction to data visualisation

1.1 What is data visualisation?

Data visualisation is the presentation of data or information in a graphical format. It is a way of

visually communicating information – often quantitative in nature – in an accurate and compelling
format.

For centuries, people have used visual representations such as charts, maps and diagrams, to
understand and communicate information. One reason for this is that the brain can process and
retain visual information more easily than verbal information (words or numbers). We see
immediately, with little effort. Thinking (cognition) is much slower and less efficient.

Data visualisations, ranging from simple line charts to elaborate interactive maps, have become an
increasingly popular approach used by researchers to communicate complex information and reach
audiences outside the research community. Effective visualisations can communicate large
amounts of information clearly, and have the potential to alter perceptions, influence people and
bring about change.

1.2 The value of data visualisations

 Visualisations can make sense of large amounts of data by revealing patterns and relationships.

 Visualisations help to communicate information and patterns to others: they can explain data
using a small space, and can be easily shared with others.

 Visualisations can tell stories in a compelling and immediate way which can be more
memorable than words.

 Effective visualisations can alter perceptions, influence people and bring about change.
The use of data visualisations to communicate public health knowledge and influence change is not
a recent phenomenon, as shown in the examples below:

John  Snow’s  map  of  deaths  from  Cholera  in  London  

In 1854 John Snow mapped deaths

from a Cholera outbreak in central
London to show its cause. The map
(marked with dots for a death, and
crosses for water pumps) showed that
cases were grouped around a water
pump on Broad Street.

As a result, the handle of the

contaminated pump was removed and
further deaths were prevented. His
work, promoting the idea that disease
can be spread through contaminated
water, was a major turning point in
the history of public health.

Florence  Nightingale’s polar area diagram of the causes of mortality during the Crimean War

In 1858 Florence Nightingale

published a diagram showing
what soldiers in the British army
died of at field hospitals during
the Crimean War. The graphic
shows the number of deaths that
occurred from wounds (in red),
from preventable diseases (in
blue) and from other causes (in
black).

Florence Nightingale showed that

the biggest cause of death was
the result of preventable diseases,
and successfully used this to
advocate to the government for
better medical and sanitary
conditions for soldiers.
2. Planning an effective visualisation
According to David McCandless (visualisation guru), good visualisations share several key features:

1. Function: they let you see trends and patterns clearly.

2. Form: they are visually appealing and well structured to attract readers and hold their attention.
3. Integrity: they portray the data accurately and honestly.
4. Interesting: they are relevant and meaningful, or reveal new information.
With the many free and low-cost tools that are now available on the internet, artistic talent is not
essential to produce a good visualisation and  interactive  visualisations  don’t  necessarily  require  
programming skills. However, good visualisations do require visual thinking and thorough planning.

2.1 Start with the objective and the audience

Identify the purpose of the visualisation:

 What story do you want to tell?
 Who do you want to reach and what does your audience care about?
 What do you want your audience to do after seeing the graphic?

2.2 Explore and select the data

 What datasets are relevant to the story?

 What are the interesting patterns and trends that should be represented and emphasised?
 Look for outliers or patterns that don’t  follow  your  expectations
 What do you think the readers will try to do with the visualisation?

2.3 Choose the right type of chart

Data visualisations will only be successful if they communicate the data easily, efficiently and
accurately, and this requires choosing the right type of chart.

The right chart should:

 Clearly indicate how values relate to one another.
 Represent the quantities accurately.
 Make it easy to compare the quantities.
 Make it easy to see the ranked order of values.
 Make it obvious how people should use the information.
Source: Stephen Few, 2013
Column/bar charts
Pros

 Clear and easy to read.

 Represent quantities easily and
accurately.
 Ordering bars by value makes comparison
easier.

Cons
 Can be dull.
 Difficult to do anything visually interesting
with shape or colour without destroying
clarity.

Pie charts
Pros
 Pie charts are good for showing
percentages.
Largest  They are most effective when there is one
segment dominant value.
starts
at 12
o’clock

Cons
 Don’t  represent  values  accurately,  which  
makes it hard to compare values with in a
chart or between charts.
 Simple or stacked bar charts may be a
more effective way to show portions of a
whole

Examples of bad pie charts

Too many values (and colours) Values  don’t  add up to 100 Too many similar values

12%
20%
65%

40%

 Don’t  show  more  than  4  values  using  a  pie  chart  (group  the  rest  into  “other”  category).
 Consider stacked bar charts as an alternative to pie charts.
3. Design tips for your visualisation

3.1 Represent the data accurately

When producing a graphic, it is important to ensure that it tells the truth about the data. It is
possible to (unintentionally) produce a misleading or ambiguous graph.

1. Always scale to zero

Bar/line charts without a zero

baseline are misleading as they
exaggerate differences
between values.

These two graphs show the

same data.

2. Don’t  use  3D  charts

3D charts distort the

33% 33% 25% 25%
data by giving
prominence to the 50%
section closest to the
front, and distract
viewers from the 33%
actual values.
3. Circle size by area, not radius

Incorrect proportion based on radius (rather than area) can lead to differences being exaggerated.

The relative size of these

pies is based on their radii.
In this example, the largest 1 2 4
circle is actually 16x the
size of the smallest circle

In this example the

proportions are correct. 1 4
2
Area = r2

3. Provide context

Lack of context can result in different interpretations of the data.

(Source: diagrams edited from Tufte, 2001)

More tips for representing the data accurately

 Use clear, detailed and thorough labelling to defeat any ambiguities.

 Write out explanations of the data on the graphic and label important events.
 For groups of graphs, use consistent scales, colours and order of categories.
 Use deflated and standardised units of money rather than nominal units.
3.2 Style the chart simply

The most important point to consider when styling the chart is that it is clear so that the viewer can
see trends and patterns easily. This usually is achieved through simplicity in design and freedom
from distractions.

Tips on styling a chart:

1. Use a limited colour palette

The human eye can only distinguish between a few

different colour shades on a single visualisation.

2. Don’t  fill with patterns.

These can create disorientating visual effects and have

no relevance. Instead, use shades of colour.

2. Mute gridlines, or remove gridlines completely if they are not necessary.

3. Label as directly as possible as this is helpful for all readers

C
A
A
B

B
3.3 Use colour with caution

The role of colour in a visualisation should be to compare and contrast one element from another,
and to highlight the important points of the graphic. Used well, colour can enhance and clarify
visualisations.

The colour wheel

monochromatic

Analogous complementary
(similar
colours)

 Monochromatic series of colours can be  Analogous colours can distinguish multiple data
used to make a heat map. series without clashing.

 Complementary colours can be used to  Bright (red) or dark (black) colours can
show highlights and detail. emphasise the important line.

20

15

10

0
Dos  and  don’ts of using colour in charts
(Adapted from The Wall Street Journal guide to information graphics by Dona M Wong, 2010)

Don’t Do
Use multiple colours to Use the same colour to Use darker shades or a
represent the same kind of represent the same variable. different colour to highlight the
data. focal point.

Don’t Do
Use complementary colours in Use graduating shades of one
a multiple bar chart as the colour or analogous colours
colour contrast is distracting

Things to consider when choosing colours

1. Colours have meaning (and these meanings vary across cultures). For instance, red is associated
with  loss  in  business  so  shouldn’t  be  used  for  positive  numbers  in  a  bar  chart.
2. Many people are colour blind, in particular they can have problems distinguishing between red
and green. Common alternatives are red-blue or purple-green.
3. Different colours (hues) that are the same lightness (gradient) can also be indecipherable for
colour-blind readers and can appear the same if they are printed in grey scale.
4. Warm colours (red, orange, yellow) appear larger than cool colours (blue, green), so red can
visually overpower blue even if used in equal amounts.

Web tools for choosing colours

ColorBrewer http://colorbrewer2.org/ Provides colour-blind safe, print friendly and

photocopy friendly colour schemes.
Adobe Kuler https://kuler.adobe.com/ A searchable database and a tool to adjust the colour
palette you are interested in.
COLOURlovers www.colourlovers.com A database of palettes that are sorted by themes.
3.4 Choose a clear font

Choosing the right font is important because it determines the tone of the graphic. For charts, the
role of the font is to describe the information clearly so it must be legible. Poor font choice can draw
attention away from the data and the message.

Which font to use?

Serif Sans-serif
 Choose a font that is easy to read (especially on a

d
computer), and with a neutral tone.
 For digital or interactive visualisations sans serif
fonts tend to be easier to read on screens than serif
fonts. Examples include: Calibri, Myriad Pro,
d
Helvetica, or franklin gothic Serif fonts have a Sans-serif
small line fonts (meaning
 For printed visualisations serif is easier. E.g.
attached to the without serif)
Georgia, Minion, Garamond, Baskerville end of the strokes don’t  have  the  
 Research has found that the use of Baskerville in a letter. line.
increases the likelihood of a reader agreeing with a
Serif fonts are Small
statement by 1.5% compared to other fonts. easier to read in on-screen text
 Comic Sans MS is more informal and not taken printed is easier to
seriously. documents. read in a
 Don’t  used  highly  stylized  fonts such as Brush Script sans-serif font.

Basic rules of type legibility in charts

(Source: Wall Street Journal guide to information graphics)

 Don’t  set  the  type  too  small  or  too  condensed.

 Use bold or italic only  to  emphasize  a  point.  Don’t  use  bold  and  italic  at  the  same  time.
 Don’t  use  ALL  CAPS  as  it  is  hard  to  read,  upper and lower case lettering is easier to read.
 Avoid knocking white type out of black or colour.
 Avoid hyphenation
 Don’t  set  type  at  an  angle

Is this easy to read?

Is this easy to read?
Decreasing
legibility Is this easy to read?
Is this easy to read?
IS THIS EASY TO READ?
Is this easy to read?
3.5 Structure the story

Structure relates to how the data is organised on the page. It guides the narrative of the piece, and
is key for understanding the visualisation/s and telling the story.

Title
As with storytelling, the beginning (title) is important. The title should make it clear to the viewer
what it is they are looking at. The sub-heading could be a key finding.

Layout and ordering

 Establish a structure which sets out the messages with a logical flow (top to bottom – left to
right).
 Organise the information to emphasise what you are trying to say to the audience: emphasise
the most important visualisation by making it the biggest object on the page, or by placing it at
the centre/top of the page.
 If the story has many sections – break it into manageable compartments with sub-headings to
guide the user through it.
 Make sure that the layout, colours and keys are consistent across charts.

Narrative commentary

Data visualisations can often be enhanced by narrative commentary. Adding text to a visualisation
can help to frame the data that is being presented, and also to guide the user through the story.

Text can also provide

Embedding focused messages additional detail or context not
in text boxes, or on a chart, can shown on a chart
be a helpful way of highlighting
important features
 4. Tools to create visualisations
 Datawrapper http://datawrapper.de/
 Infogr.am http://infogr.am/
 IBM Many Eyes http://www-958.ibm.com/software/data/cognos/manyeyes/
 Visual.ly http://visual.ly/
 Google Fusion Tables https://support.google.com/fusiontables/answer/2571232
(for maps)
 Tableau Public http://www.tableausoftware.com/

Blogs and resources on data visualisation

 Gapminder www.gapminder.org
 Guardian Datablog http://www.theguardian.com/news/datablog
 Information is Beautiful www.informationisbeautiful.net
 On Think Tanks http://ttdatavis.onthinktanks.org
– On Data Visualisation
 Visualising Data http://visualisingdata.com
 Worldmapper www.worldmapper.org

Books, articles and references

Information is Beautiful The Visual Display of Facts are Sacred by The Wall Street
by David McCandless. Quantitative Simon Rogers, 2013. Journal Guide to
Second Edition, 2012. Information by Edward Information
Tufte, Second Edition, Graphics by Dona M
2001. Wong, 2010

 Making visualisations: a survival guide by Gregor Aisch.

http://www.slideshare.net/vis4/making-data-visualizations-a-survival-guide
 Data Visualisation for human perception by Stephen Few, 2013.
http://www.interaction-
design.org/encyclopedia/data_visualization_for_human_perception.html
 Visualizing Information for Advocacy – An Introduction to Information Design
http://www.opensocietyfoundations.org/reports/visualizing-information-advocacy-
introduction-information-design