Presentation of Tables,

Graphs and Maps

Alex Thomson

December 2020

Access our full collection of guides, materials and more at:

stats4sd.org/resources

©2020 Statistics for Sustainable Development

Except where noted, this work is licensed under a Creative Commons Attribution-
ShareAlike 4.0 International licence.

This document should be cited as:

Thomson, A (2020). Presentation of Tables, Graphs and Maps. Statistics for
Sustainable Development, Reading, UK

Stats4SD
 Presentation of Tables, Graphs and Statistics

Contents

Introduction ................................................................................................................ 3

General Guidance ...................................................................................................... 4

Considering the Message ......................................................................................... 7

Tables ........................................................................................................................... 9
General Guidance ................................................................................................................. 10
Statistical Information .......................................................................................................... 17

Graphs ....................................................................................................................... 17
Choosing the right graph..................................................................................................... 19
General Guidance ................................................................................................................. 28
More Examples ...................................................................................................................... 39

Maps ........................................................................................................................... 44
Map Types ............................................................................................................................... 45
Examples: ................................................................................................................................ 47
Using colour and categorising values ............................................................................... 48
Map Elements ........................................................................................................................ 49
Example ................................................................................................................................... 51

Accessibility ............................................................................................................... 53
Tables ....................................................................................................................................... 53
Graphs/Maps.......................................................................................................................... 54
Colour ...................................................................................................................................... 54

Advice on Software .................................................................................................. 59

Checklist..................................................................................................................... 60
General Guidance ................................................................................................................. 60
Tables ....................................................................................................................................... 60
Graphs ..................................................................................................................................... 61
Maps ......................................................................................................................................... 61

Other Resources ...................................................................................................... 63

Page 2 of 64
 Presentation of Tables, Graphs and Statistics

Introduction

Visualising data is an essential part of communicating messages and results to any

form of audience. An ineffective visualisation of data can communicate a very
misleading message.

Building skills in data visualisation can help you to understand and see important
results in other people’s tables, graphs, and maps. This is in addition to enabling
you to create informative visualisations of your own.

The aim of this document is to provide comprehensive guidance on the

presentation of data in tables, graphs, and maps. This will include both general
guidance and more specific advice on different types of visualisations. We intend to
provide some principles of good graphical, tabular, and cartographic practice. By
providing this advice, we hope to assist anyone in their future work, especially when
it comes to the writing up of research results for an audience.

This guide is intended for anyone who wishes to develop their data visualisation
and reporting skills. The advice presented here will be applicable to a wide variety of
situations and is not specific to certain topics. Additionally, we hope that users of all
ability levels will be able to take this advice to mind in their future projects and their
everyday interactions with data.

This resource will start by exploring some general guidance on the presentation of
data before going into more specific detail on the use of tables, graphs and maps
(an increasingly popular method of presenting data). It then provides advice on
ensuring your visualisations are accessible, with consideration on the use of colour.

Page 3 of 64
 Presentation of Tables, Graphs and Statistics

General Guidance

Through developing your data visualisation skills, you can generate a wide variety of
graphical/cartographic/tabular representations of data. This could be anything from
simple bar graphs and line graphs to complicated cartograms. However, regardless
of the complexity of your chosen data visualisation technique, there are certain
principles that should always be followed:

• Effectiveness

By effectiveness, we mean you should be ensuring that you are using the right
type of visualisation for your objectives and priorities. This is the first crucial
step in making sure what you produce is effective at displaying the message you
intend to show. If you pick the wrong method, your visualisation will not be
effective regardless of its quality.

Maps are of course for displaying data which have some form of geographic
component.

Tables are suited for presenting structured numerical information; consider

tables of means across some groups, frequencies, or some statistical
information. This makes them ideal for when the message is in the specific
numbers and potentially the relationship between them.

Graphs are quite multi-purpose; there is a type of graph for almost any
message you could be wanting to convey. In general, we would choose to use
them for indicating trends, making broad comparisons, or showing
relationships.

• Readability

All elements of your visualisation should be legible, understandable, and

coherent. In a word, readable. While this largely relates to any textual elements
of your visualisations, the principle is applicable to the whole visualisation.

This includes having titles and headings which concisely explain the content. It
should be informative without being overly long and confusing. The same goes
for any further labels such as axis labels for graphs, column headings for tables
and geographic labels on a map.

Details to consider mentioning include measurement units, geographical

coverage, time, the source of the data and any relevant statistics.

Page 4 of 64
 Presentation of Tables, Graphs and Statistics

Of course, the elements of a visualisation will vary depending on what visual you
produce, but they should always be easy to read and understand. You can
achieve this by avoiding language beyond the scope of your target audience,
providing the necessary information needed to read your visual and presenting
the element in a simple and tidy manner. You will find further guidance on
specific elements in each of the subsequent sections of this guide.

• Tidiness

A visualisation should never be cluttered. This follows on from readability,

although more specifically relates to positioning and spacing of elements as well
as avoiding using unnecessary elements.

This includes making sure no elements are overlapping; there should be

adequate spacing between them without there being so much that it makes the
visualisation look empty. This can also be described as making good use of
“white space”.

There is much more to be said on this topic, but these are mostly specific to the
type of visualisation you are using. The general principle of ensuring your visual
is neat and organised is always applicable.

• Accessibility

Accessibility has become an increasingly important aspect of data presentation

in recent years. Ensuring good practice in accessibility will help in getting an
even wider audience to see our research and use our results. The Government
Statistical Service makes content accessible to those with impairments to their
vision, hearing, mobility, and thinking/understanding skills. For our purposes, we
are mostly concerned with visual impairments.

There are some general principles on accessibility, including making sure you
explain any uncommon abbreviations, avoiding clutter and keeping information
concise. However, we are focusing on the use of colour. Further guidance on
this is included in the accessibility section of this document. This includes
considerations of colour blindness, cultural context, and the use of
saturation/hue/luminance.

• Consistency

This is mostly relevant when you are intending to use multiple visualisations
across your report. When doing so, it is important to ensure you maintain a
level of internal consistency.

Page 5 of 64
 Presentation of Tables, Graphs and Statistics

This involves many aspects. For instance, if you intend to disaggregate your
visuals by the levels of a variable, pay attention to the order you put these
categories in. They should be kept to a logical or ascending/descending order
and this order should be kept the same for the sake of consistency and
readability.

The same goes for when using colours to indicate certain characteristics of the
data; keep the meaning of the colours consistent.

Of course, this is also important for all the smaller details such as the font, size
and face (bold/italic) of text. In essence, try to keep the formatting between
visualisations as similar as possible. Generate your personal visual style and
stick to it. Changing things up too much will just confuse your audience and
reduce your visual’s readability.

• Informative

A good data visualisation serves to succinctly show a message about our

findings. We aim to inform our reader. Usually, it would be accompanied by
some text which helps to interpret the visualisation, placing it into a wider
context or providing more formal details such as the results of a relevant
statistical analysis.

However, a good data visualisation should be self-explanatory and should be

able to serve as a stand-alone piece. The reader should be able to understand
the message without constantly referring to the text. Much of this can be
accomplished by sticking to the particulars of keeping your visual tidy and
readable.

Whenever creating a table, graph, or a map, you should include the source of
the information from which the visualisation was created. This aids the
credibility of your visualisation but also ensures a properly informed audience.
An exception is when all information that is used for visualisations in a report
comes from the same source. In this case, you should clearly indicate the
source in advance of your visualisations.

This also means making sure that your visual is necessary in the first place.
Consider the following: Can you achieve the same message with some simple
text? Can a visualisation accurately demonstrate your results, or would it be
distracting? Are your results too complex to visualise in isolation?

These six principles are relevant regardless of which visualisation you choose to
create. In the following sections you will find guidance that is more specific to tables,
graphs, and maps. While the guidance is specific to the different forms, they all tie
into the central principles described here in this first section.

Page 6 of 64
 Presentation of Tables, Graphs and Statistics

Considering the Message

Two of the above principles go beyond the specifics of what you put into your data
visualisations: effectiveness and being informative. Consideration of these two
principles does not start when you plot your variables. They are principles which
should guide your entire research process, including presentation.

This guide will make regular reference to considering what is appropriate for your
message, your results and your purpose. For your data visualisations to be effective
and informative, you need to think hard about the message you want them to
convey. This will often come back to an original research question. These research
questions should always be guiding you in the creation of data visualisations.
Effective data presentation needs to have something to say, and what it says should
be relevant.

Consider this as a process:

1. We start with our research questions that we want to help answer through
our research.
2. We can break these up and consider how we will answer them. What are
going to be the key points we will need to investigate to answer these
questions?
For example, say we want to research the prevalence of a disease across
areas within a country. We can decide that we are going to need to make
points about the overall prevalence, the geographical variation, the
explanations, compounding variables. We could look at these as the building
blocks of our messages. Our messages are what we want people to
remember and they will all stack up to help answer our bigger questions.
3. After this, we conduct our analysis and pick out our key findings. These key
findings will similarly be informed by our existing research questions and
pre-conceived ideas about what our messages will be. However, they should
always be flexible; an unexpected result should not be ignored.
4. We now need to update our messages based on what we have observed.
Our messages should always strive to be important, relevant, and interesting.
Also consider novelty; repeating a message we have heard many times over
and over will not result in a very interesting data visualisation.
5. These updated messages and key findings will inform the creation of our
presentable data visualisations. These visualisations along with our messages
help to answer our initial research questions.

Page 7 of 64
 Presentation of Tables, Graphs and Statistics

Therefore, we think about being effective and informative throughout the research
process. If your messages and questions are not effective or informative then you
cannot expect your visualisations to be.

Bringing effectiveness and informativeness into your data visualisations requires

careful consideration of the messages you have drafted. This leads to questions you
will need to ask of yourself, including:

• What variables should I include?

You should not be including more variables than are necessary. Think about the
specifics of your intended message and only include the variables which are
relevant and necessary for effectively showing this message. You should also
avoid including variables which are uninformative. If adding in a variable does
not add any explanatory value, then drop it from your visualisation.

• Which variables should I split by?

Disaggregating your findings by certain groups is a common practice. What

variables you use to do this splitting should largely be informed by your
research questions and messages.

Consider the example of geographical variation of disease prevalence. Explicitly,

we know we will need to look at how our results vary by geography, but we may
also want to consider variables which could help to explain the geographical
variation. So, we could split by levels of economic deprivation or rurality. It
always comes back to keeping relevant and important.

• Which graph/table/map should I use?

This involves thinking carefully about the type of message you want to show.

Does your message mostly concern changes over time? If so, then a table
probably is not suitable unless it is quite a short time frame with few points.

A few graphical options would be suitable including line graphs, column charts,
slope charts etc. From here, the choice would now be dependent on the types
of variables you want to show. Concerned with totals? Then consider column
charts. Concerned with averages/rates? Consider a line chart.

Deciding on the right type of graph is a process, starting from your overall
message and working down:

1. What is the purpose of your message? (Change over time?

Distribution? Spatial? Correlation? Etc.)

Page 8 of 64
 Presentation of Tables, Graphs and Statistics

2. What variables are you plotting and how many? (Categorical?

Continuous?)
3. How much data needs displaying? A lot or a little?
4. What are the measurement units? (Averages? Totals? Rates?
Proportions?)

Considering these questions should help you to narrow down what type of data
visualisation is most appropriate. There are too many possible variations to
consider here, but these points should help guide your thinking process.

Tables

When thinking of data visualisations, tables may not be your first choice as they are
not as visually remarkable as a graph. However, tables are a crucial tool in
presenting data and results as they have the advantage of much greater specificity
than graphs and are usually simple to understand. Generally, it is harder to read
patterns in tables than in graphs. Therefore, graphs should be used when you want
to focus on patterns, trends and relationships that do not necessarily require the
exact values to be understood.

A table would therefore be more appropriate than a graph or map if:

• You are asking the audience to compare individual values directly

• You are wanting to include both the values and some derived measures such
as percentages or indices. These are harder to show succinctly all together
on one graph.
• You want to include summary statistics such as means or totals
• You need to show values with very different magnitudes together.
• If users may want to use the data for their own analysis or reference.

Reference tables contain extensive information for people to look up.

• They are useful for archival purposes rather than analysis.

• They should include detailed metadata about the information presented:
what, where and when of the data.
• They usually appear as appendices.

Demonstration tables are probably what you think of when we mention tables for
research purposes.

• They are intended to reinforce a point by showing statistics or values that

can be quickly assimilated by the reader.

Page 9 of 64
 Presentation of Tables, Graphs and Statistics

• They are included within the text to allow readers to follow the general
argument and without having to flip back and forth to refer to the relevant
information.
• It is important they are clear and well-presented, usually using reasonable
approximations to keep figures to a few significant figures.
• Very large demonstration tables can be confusing and intimidating. If all the
information is truly required, it should be split across multiple smaller tables.

The following guidance mostly concerns the formatting of demonstration tables

although the general principles are applicable to both forms.

Reference tables however are not designed to draw attention to specific numbers,
patterns, or comparisons and therefore advice on topics such as ordering of
columns and rows are not especially relevant.

General Guidance

Title, column headers and labels

Titles and labels are very important to the design of a table as they help users
understand what is being presented. The titles and labels make sure the table
works on its own and can be read within a different context than its original
presentation.

You should consider including the following information in tables within either titles,
labels, headings or possible footnotes, the choice of which points depends on your
data and how important the details are to understanding the information:

• Analysis units (people, households etc.)

• Types of statistics (totals, means etc.)
• Units (thousands, kg, $)
• Geographical coverage
• Time period
• Source of data
• Key quality information

Page 10 of 64
 Presentation of Tables, Graphs and Statistics

The table below (1a) contains many errors, some of which we will come back to
shortly. However, let us firstly address its labelling issues.

• It is not clear what the table means by diet and what the difference between
the levels of diet are.
• These labels are very uninformative and fail to be specific. We should not
have to look to footnotes to define the contents of a column.
• The measurement units have not been specified.
• The title does not effectively stand out compared to the rest of the table

Table 1a: Mean growth rates and water intake across four diets.

Diet1

Variable I II III

Growth rate 89 145.32 97.128

Water intake 108.4 121.29 121

1Diet I = Control; Diet II = Alternative

Let us go step by step and fix these labels and titles first. We have:

• Centered and put the title in bold face so that it stands out and specified a
better meaning for diet.
• Renamed diet to supplement as this is a more appropriate heading.
• Directly named the supplements rather than through footnotes.
• Included the units of measurement for the variables so the numbers can be
understood.

However, we can still make further improvements.

Table 1b: Mean growth rates and mean water intakes for four dietary supplements

Supplement

Variable None Lucerne Sesbania

Growth rate 89 145.32 97.128

(g/day)

Water intake 108.4 121.29 121

(ml/kg0.75)

Page 11 of 64
 Presentation of Tables, Graphs and Statistics

Comparing numbers (rounding, decimal places, and alignment)

Tables will require the reader to compare numbers. If these numbers are differently
rounded or contain differing levels of significant figures/decimal places, then
comparing them becomes more difficult. Here are some things you can do to make
this process easier:

• The same level of precision should be used within each variable. The
precision can vary between them, because different measures or ranges will
require different levels of precision to make an accurate comparison, but it
should be consistent within each variable.
• It is best to minimise the number of decimal places such that comparisons
can be effectively made without any loss of information.
o It is uncommon to need more than 3 decimal places. If you are dealing
with incredibly small precise values, then consider using scientific
notation. However, bear in mind that scientific notation is harder to
understand for most readers.
• Rounding larger numbers is also advisable depending on your purpose.
Demonstration tables usually use suitably rounded numbers that effectively
illustrate the message. Reference tables tend to use a higher level of
precision as users typically require a more exact number.
• Using commas to separate large numbers can make these numbers more
easily readable, although if you are presenting large numbers you should
also consider standardising the numbers into thousands, millions etc.
• Generally, numbers should also be right aligned, as should the column
headings. The decimal point should line up.
• Decimal numbers between 0 and 1 (or 0 and -1 if negative) should start with
a 0 and not a decimal point.

With this guidance in mind, we can see that in our example we still have some
issues:

• We are inconsistent on how many decimal places we use within our

variables, making comparison difficult.
This is especially true between water intake for Lucerne and Sesbania as the
mismatched precision means we cannot tell which is larger.
• We have not aligned our numbers correctly as the decimal points do not
match up with each other.

Page 12 of 64
 Presentation of Tables, Graphs and Statistics

Table 1c: Mean growth rates and mean water intakes for four dietary supplements

Supplement

Variable None Lucerne Sesbania

Growth rate 89 145 97

(g/day)

Water intake 108.4 121.3 121.0

(ml/kg0.75)

We have helped alleviate these problems by:

• Ensuring our level of precision is consistent within our variables.

• Minimising this precision to the level required. Adding decimal places to
growth rate increases the precision of the individual values but is not
required for comparison. While, as two of our groups both round down to
121 for water intake, we need at least one decimal place to make a proper
comparison.
• We have right aligned our numbers and column headers.

Orientation

A table’s orientation can significantly affect its readability. It is much easier to

compare numbers within a column than within a row. Therefore, if we intend to
compare numbers across groups according to several variables, the variables
should define the columns and the groups should define the rows.

This is true of both demonstration and reference tables.

In our continuously improving example, we are using variables to define our rows.
This is making it more difficult to compare between the groups. Here is the updated
version:

Table 1d: Mean growth rates and mean water intakes for four dietary supplements

Supplement Growth rate Water intake

(g/day) (ml/kg0.75)

None 89 108.4

Lucerne 145 121.3

Sesbania 97 121.0

Page 13 of 64
 Presentation of Tables, Graphs and Statistics

Order of rows and columns

Another way to improve the layout is to consider the ordering of rows and columns.
If there is some logical ordering to the groups, maybe because it is an ordinal
variable, then you should keep them in this order. However, if there is no logical
order, it is advised to order them according to the most important variable. The
most important variable will depend on your data and objectives.

Additionally, in cases when one of your groups is “none” or “other”, it is often

sensible to put these as the bottom rows. A “none” group often serves as a useful
baseline to compare all other groups against. An “other” group is usually a
combination of rare instances and lacks specified information so is rarely useful to a
table’s overall message.

Ordering of rows and columns is generally not so important when creating

reference tables, as these are less likely be used for comparisons or to spot
patterns. However, keeping the rows in some form of logical order will likely still
help with the table’s readability.

In our example, we can keep our columns in their current order as there is no
natural reason put one before the other. However, our rows are not ordered and
therefore we can still make comparisons easier by ordering them by growth rate.
Conveniently, this puts our “none” row last.

Table 1e: Mean growth rates and mean water intakes for four dietary supplements

Supplement Growth rate Water intake

(g/day) (ml/kg0.75)

Lucerne 145 121.3

Sesbania 97 121.0

None 89 108.4

Borders

Borders should be using sparingly and only when necessary. They can be used to
help separate parts of a table or groups of rows. However, using them too much
just makes the table look cluttered and can interrupt numerical comparisons.

Therefore, borders should be avoided within the main body of the table and there
should be no vertical lines. Horizontal lines should only be used to separate out a
table’s header and footer from the main table body and the page itself. Horizontal
borders are effective between column headers if there is a hierarchical grouping
between them.

Page 14 of 64
 Presentation of Tables, Graphs and Statistics

In our example, we have put borders around every single cell. This is completely
unnecessary.

In table 1f, we have improved our table by removing all vertical lines. These are not
needed since there is enough white space between the values and they have been
properly aligned, making the columns appropriately distinct.

The only borders we have kept are those that separate the table from the page and
one that splits the column headers from the table body.

Table 1f: Mean growth rates and mean water intakes for four dietary supplements

Supplement Growth rate Water intake

(g/day) (ml/kg0.75)

Lucerne 145 121.3

Sesbania 97 121.0

None 89 108.4

We now have a table which is tidy, readable, effective, consistent, and informative.
However, there are some other areas of advice which may be useful.

Font

Be consistent with your font and ensure it is professional. It is recommended to use

sans serif fonts such as Open Sans, Arial, Helvetica, Tahoma, or Verdana. Bold
should only be used for titles and headings. Keep changes in font size to a
minimum and avoid small fonts.

Grouping of rows and columns

Grouping is often useful to maximise the amount of information displayed while

maintaining the table’s effectiveness and readability. For example, you may use
levels of more than one categorical variable to define your rows or you may wish to
present more than one measure for a variable (such as wishing to show the mean,
the sample size, and the standard deviation).

Below is an example blank table that may be created for such a purpose.

While there may be a temptation to put horizontal borders between the different
groups of rows and vertical borders between the groupings of columns, this should
be avoided. Instead, using white space between the groupings is a much neater
alternative that effectively separates out the information and keeps focus within the
groups.

Page 15 of 64
 Presentation of Tables, Graphs and Statistics

Table 1g: Skeleton table example for grouping of columns and rows

Growth rate Water intake

(g/day) (ml/kg0.75)

n mean sd n mean se

Supplement

Lucerne - - - - - -

Sesbania - - - - - -

None - - - - - -

Site

Site 1 - - - - - -

Site 2 - - -

Overall - - - - - -

Summary rows and columns

Summary rows and columns are quite useful for providing extra information that
may be useful for interpretation. These should be placed at the bottom or right of
the table unless they are the primary message of your table, in which case putting
them first and then disaggregating is acceptable.

Other

Some other general tips include:

• You can use footnotes to provide additional contextual information including:

o Source
o Units of measurement
o Statistical information (such as level of significance)
o Any mitigating information that helps with interpreting the figures.
• If the table spans multiple pages, include the table’s heading at the top of
each page.

Page 16 of 64
 Presentation of Tables, Graphs and Statistics

• Do not put the table in the middle of text. Ensure an effective and neat layout
between your table and your text.

• If your table would only need two or fewer columns and a handful of rows,
consider just writing this information out in text.

Statistical Information
In more formal scientific papers, it is often required to include statistical information
within tables and present the results of statistical analysis. This following tips offer
guidance on presenting descriptive statistics, measures of precision and identifying
statistical significance in tables:

• In tables, we often want to present useful summary statistics including the

mean, the sample size, and the standard deviation.
• A sample size is usually presented in its own column which can be called “n”
if we are addressing a statistically literate audience.
• A standard deviation is commonly included inside parenthesis after the
mean or sometimes within its own column. The choice may depend on the
number of columns already being used.
o If you require standard deviations across multiple columns of means,
then giving each their own column and putting the deviation in italic
should make it stand out more (see table 1g). This will also be neater
and more readable than using lots of parenthesis.
• Measures of precision such as standard deviation or error are usually
presented with one more decimal place than the mean, although this is not a
strict rule.

Graphs

While a table can be very effective at displaying a clear and interesting message, a
graph can deliver this message more succinctly and provide greater visual
emphasis.

It can sometimes be difficult to spot the pattern in tables even if it is well formatted
and accompanied with clear explanations. However, with a graph, we can make this
pattern the focal point.

Additionally, we can visualise the entirety of our data, while tables tend to be limited
to summaries or statistical results. Therefore, graphs allow us to highlight a much
greater range of messages about our data. Graphs can potentially reveal insights
about our data and results that would be hidden by a table.

Page 17 of 64
 Presentation of Tables, Graphs and Statistics

As with all visualisations, a good graph should make sense when seen out of
context, with its message intact. A poorly formatted graph can, whether
intentionally or not, mislead its viewers into thinking patterns are more/less
significant than they are, especially when taken out of context. Regrettably, we often
see graphs being intentionally used to mislead and we should be sure to avoid
committing the errors that such graphs commit.

For example, the graph below has become one of the most notorious recent
examples of a misleading graph:

Generally, we would expect the y axis to start with 0 at the bottom and increase
upwards, not the other way around. At first glance, it appears there was a significant
drop in gun deaths after the introduction of “Stand Your Ground” laws. However,
upon reading the y axis labels, we can see that the truth is the exact opposite.

Page 18 of 64
 Presentation of Tables, Graphs and Statistics

Choosing the right graph

Choosing the right graph to fit your message is arguably the most important step in
creating effective data visualisations. As mentioned in the section on Considering
the Message, the first question to ask yourself is ‘what is purpose of the message I
want to show?’

A good starting point is to consider if your message focuses on any of the following
purposes:

• Highlighting the distribution of a variable

• Highlighting change over time
• Highlighting a relationship between variables
• Ranking categories by a variable of interest
• Comparing values across groups
• Exploring a part to whole relationship
• Showing a geographical pattern
• Exploring deviation

Different plot types have different purposes. Some will have the same purpose but
show it in a different way or are useful for different kinds of variables. For instance,
a line chart and a column chart are both suitable for showing changes over time,
but the latter is more suited to totals while the former is better for rates and
averages.

Some plots can have multiple purposes. For instance, a stacked bar chart can show
both the magnitude of numbers comparatively and a part to whole relationship,
depending on the scale used.

Graphs and purposes are not mutually exclusive and the list presented below is not
exhaustive. Other graph types do exist, and you may have other types of messages
in mind from your data analysis. This section offers an overview of some key
purposes of messages as well as graph types

Page 19 of 64
 Presentation of Tables, Graphs and Statistics

Distribution

The purpose of distribution graphs is to show the variation or shape of a continuous variable within your dataset to either show
the frequency or distribution of all values in your data, or to highlight key measures of the distribution i.e. the range, median,
maximum, and minimum. They are useful ways of highlighting lack of uniformity or equality in the data.

Histograms – The standard graph for this Boxplots – Summarise the data using a box to Violin plots – Similar to boxplots except they
purpose. Uses bars to show the distribution of represent the inter-quartile range and lines to show the entire distribution of the data, making it
the data with the bars representing frequency. represent the median, the maximum, and the easier to spot specific spikes and crowding of
Keep the columns tight together to keep the minimum. Outliers are often presented observations. Effective with data with more
users focus on the shape. separately as dots. Useful for comparing complex distributions.
multiple distributions.

Dot plots – Very simple graph designed to Jitter plots – Visually similar to scatterplots.
show the range of data. Dots representing Each individual observation is plotted using a
the minimum and maximum are connected point with the continuous variable on one axis
using a line. Useful for a non-statistically and a categorical on the other. A slight random
literate audience. variation is applied to avoid overplotting.

Page 20 of 64
 Presentation of Tables, Graphs and Statistics

Some less common distribution graphs:

Dot strip plots – Show individual values in Barcode plots – like dot strip plots but uses Population pyramids – Essentially back-to-
a distribution across a horizontal line. Can vertical lines instead back histograms. Normally reserved for
be difficult to read when too many values breaking down a population by age and sex.
are the same

Page 21 of 64
 Presentation of Tables, Graphs and Statistics

Change Over Time

In some cases, we want to show how variables have changed over time, often to emphasise trends or patterns.

Line graph –The most common way to present time Column – Using columns to represent Slope - A simple line plot connecting
series data. Very versatile, they can also be used for values within given period. Quite useful if just a few points. Useful for showing
other forms of data, including distributions, ranks, the variable of interest is a total. Usually changing data so long as the pattern
and proportions. best for shorter single time series. can be concisely summarised into
just 2 or 3 points.
• The data can be split into multiple lines by some • Can be combined with a line to show
other variable or the lines are separate variables relationship between an amount (the
altogether. column) and a rate (line)
• It is acceptable to break the y axis on a line graph
so long as this is clearly marked.

Area chart – Useful for showing how Streamgraph – Similar to area charts but Calendar heatmap – Often each box will present
contributions to totals by different categories designed to showcase changes in proportions a day, week, or month, depending on your time
scale. Each box is coloured according to the value
change over time as well as the trend of the between different categories over time. Like area
of another variable or frequency of observations.
of the total overall. However, can be difficult charts, they are visually striking, and the general
Useful for keeping track of recent trends or trends
to read all but the bottom category. patterns are easily readable but exact values are
over a short period of time.
likely not of importance to the audience.
Page 22 of 64
 Presentation of Tables, Graphs and Statistics

Magnitude

Another common reason to create a graph is to compare values, usually frequencies/totals within groups. Many of the following
graphs can also be used to plot values such as comparing means/totals between different groups as well as within groups.

Bar chart – One of the most standard graphs used in data Clustered bar chart – allows you to show multiple series of bars
visualisation. Simple to use and can be quite versatile. Can use either by splitting the categories into subgroups. For instance, using
vertical or horizontal bars. Horizontal bars often useful when you countries across your x axis and then splitting the country bars
have lots of categories or long category labels that do not fit well into separate bars for male and female.
under vertical bars.
• Choice of which variable to use on the x axis and which to
• The gap between bars should be slightly narrower than the use as sub-groups will depend your purposes as it is easier
width of a single bar to compare within groups and the general pattern rather
• You can add value labels to a bar chart although if you do, a than across subgroups along the x axis.
table should be a valid alternative.

Page 23 of 64
 Presentation of Tables, Graphs and Statistics

Part to Whole

It is very common to want to show a proportional breakdown for each level of a categorical variable, sometime split across
another categorical variable such as location or gender.

Stacked column or bar chart – Similar to a paired Pie chart – Commonly used chart Donut chart – Same as a pie chart but
bar/column chart but the groups of bars are stacked with segments of a circle being with a space in the middle that can be
on top of each other rather than next to each other. representative of the proportion of used to provide additional information
the whole the category accounts for. such as a total.
• Either can have the chunks equal the frequency
and therefore the whole bar is equal to the total
frequency.
• Alternatively, use individual chunks adding up to
100% so the chunks represent the proportion it
contributes to the total.

Grid plot – A series of boxes, usually Tree plot – A square separated into
100, with the amount coloured in a multiple boxes to visualise a part-to-
certain colour being equal to the whole relationship which is hierarchical
proportion of the total that category while also showing the frequencies of
accounts for. each level of the categories. Can be hard
to read and designed for a more
statistically or data literate audience.

Page 24 of 64
 Presentation of Tables, Graphs and Statistics

Correlation

Often, we wish to demonstrate the relationship between two or more variables. It is quite simple to do this with basically any
type of graph, usually by using colour to denote some extra categorical variable or creating separate lines/bars. There are
however some graphical methods which can be used to prioritise the visualisation of this relationship. As previously mentioned,
combining a column and a line graph is a good way to show relationship between a total and a rate over time.

Scatterplot – Standard method for Bubble – Like a scatterplot but the points are XY heatmap – Equivalent to a contingency table
demonstrating the relationship between sized according to a third continuous variable. with categories of one variable along the y and
two continuous variables with a point Probably best not to also use colour for a fourth categories of another along the x and a square
representing each observation. variable as this will make your graph cluttered representing each possible combination.
and hard to read. Coloured to represent frequency.

Page 25 of 64
 Presentation of Tables, Graphs and Statistics

Ranking

Sometimes an item’s position in an ordered list is more important than the absolute value. Often this may also be a secondary
message from our data and therefore we can often use the same graphs as we use for other means.

Ordered bar/column charts – Essentially just a simple Lollipop chart – Similar to the above but draws more
bar/column chart but make sure the bars are ordered attention to the value as well as the rank.
according to the value they represent.

Bump – Used to show changes in ranks over time with a line Slope – You can also use the slope chart to show
per category moving up or down, crossing over with other lines changes in ranks over time across different categories.
to demonstrate the ranking of the categories at each point in
time. Avoid using too long a time scale.

Page 26 of 64
 Presentation of Tables, Graphs and Statistics

Deviation

Deviation graphs are used to emphasises variation from a fixed point of reference such as 0, a target value, or the overall/long-
term average. Also, useful to show the results of Likert scale questions. Another way to incorporate this onto virtually any plot
point would be using horizontal/vertical lines of reference to aid comparison.

Diverging bar chart – A simple bar chart Diverging stacked bar chart – Usually used Spine – Essentially two back-to-back bar
but usually presented with horizontal bars for presenting Likert scale results with the charts with the charts split by a contrasting
with values being able to take both negative middle representing the number/proportions category. For instance, the left side of the
and positive values. of observations in the neutral category. bars representing females and the right
Negative categories are then below this group side being male.
and positive categories are above.

Spatial Data

Sometimes our data has a spatial component, and we may wish to plot the geographic relationship. If there are only a few
geographic locations, a table or a bar graph may be sufficient. Consider creating a graph/map if you have lots of locations or the
geographical pattern in the data is the most important message

See section in this document on map types.

Page 27 of 64
 Presentation of Tables, Graphs and Statistics

General Guidance

Tidiness

A graph can easily become cluttered and messy. It can be very tempting to add
more information to explain more of our data or to find more patterns. However,
you should always use restraint. Focus on the key messages and do not try to show
too many all at once.

A good graph should have 1 or 2 key results to show (usually the general pattern or
relationship). It is acceptable for other minor messages to be visualised, so long as
they do not deter attention away from the focus and do not require significant extra
effort to plot.

Some key tips to keep your plots tidy include:

• Avoid using “Junk” features.

These are extra visual features which serve little to no purpose in aiding the
data visualisation other than formatting.

• Shaded backgrounds
• Borders
• Boxes around legends
• Patterns, textures, and shadows
• 3-D graphics
• Data makers on line charts (unless used to annotate specific key data
points, in which case do not mark every single point)
• Thick or dark gridlines

• Avoid Overplotting

Overplotting occurs when data points or labels overlap, making it difficult to

read individual points. Usually this is because there are too data points with the
same/similar values, or because there are only a limited number of unique
values. It can lead to graphs that are difficult to interpret or show misleading
results.

You can avoid overplotting by:

• Reducing the size of the points

• Showing just a subset of the data
• Changing the plotting symbols

Page 28 of 64
 Presentation of Tables, Graphs and Statistics

• Using transparent symbols – this is the preferred option as the areas

with lots of points become clearly darker than the rest of the plot.
• Jittering – adds slight random movement of the data points around their
true value. Makes it easier to read points where there are only a small
number of values.
• For more information, see https://www.displayr.com/what-is-overplotting/

• Do not use too many categories

The exact guidance on this will depend on your data and thus the type of plot
you want to make, but the key points are as follows:

• Do not plot too many lines on your line graph.

A key advantage of a line graph is that we can plot multiple lines (either
split by a variable or another variable entirely). However, we do not want
to use too many because otherwise we cannot follow each line
independently.
• Similarly, with any plot with splits something by a categorical variable,
avoid using too many categories. Again, this can make your graph
incredibly large and make it difficult to compare specific groups.

In the graph below, there are 27 violins placed on top of one another and we
can barely tell anything about one, let alone compare them.

Page 29 of 64
 Presentation of Tables, Graphs and Statistics

• Do not use too many variables

When we conduct an analysis, we often want to account for as many variables

as possible. Consider how many variables we may include in a model to explain
some other key indicator. However, we want to avoid doing the same for our
graphs; we only have so many axes and ways to visually split our data.

Just because we have an x axis, a y axis (sometimes 2 of them), a z axis (not

recommended), colours, sizes, shapes, line types etc. available to us for plotting
data, this does not mean we should try to use them all at once

Only use as many axes/dimensions as needed to remain informative. If you can

present your information using just one dimension, then do not use a second. If
can be done in 2, do not use a third. If you need more than 3, consider
separating them over a series of graphs.

Consider the graph below. So many dimensions have been used that the graph
has become far too noisy and can barely be understood.

Page 30 of 64
 Presentation of Tables, Graphs and Statistics

A better option would be to stick to just the 2 or 3 axes/dimensions that are most
important to your messages. In this case, let’s focus on the relationship between
the x and the y axis and how countries differ between regions (the latter
represented by colour).

Title

See previous general guidance. (pg.10)

Source

If it is likely that your graph could be shared externally or repurposed by others,

include a source note in your graph so that others are properly informed about the
data’s origin and the context of the graph.

Scales

When plotting variables, both continuous and categorical, we need to consider how
are going to go about it. What dimensions and axes are available to us? Which of
these should we use? How should we construct them?

Page 31 of 64
 Presentation of Tables, Graphs and Statistics

All variables will be plotted on some form of scale and getting this scale right is very
important to ensuring that your graph is effective. An ineffective scale can make a
graph unreadable or, at worst, misleading.

• Axes

Axes can be used inappropriately to show patterns that are not there or inflate
existing differences Therefore, you need to ensure that your choice of scale on
your x and y axes are not misleading and authentically represent your data.

For any axis:

• Label the axes appropriately so that it is clear what the graph is showing

For continuous axes:

• Use equally sized breaks along the scale

• Label every key break, axes can become quite cluttered if every single
break is labelled
• Do not leave large gaps in your axes
o For example, if you create a line graph with time on your axis but
there is some period in the middle with no data, do not just leave
a gap in your axis.
o Keep the axis together. If the gap is particularly long you could
consider an axis break.
o You should not connect the lines either side of the missing period
as this is misleading about the actual trend of your data.

On a bar chart, you should:

• Always start your axis at 0.

• Never start your axis at anything else, and do not break it either
o The height of the bars relative to the axis gives a clear indication of
the magnitude of the category or value.
o If the axis does not start at 0 or is broken, then the height
becomes misleading and differences will be inaccurately
visualised.

Page 32 of 64
 Presentation of Tables, Graphs and Statistics

Compare the next two graphs. One starts the x axis at 80% while the other
starts it at 0%. The first makes Norway look almost 4 or 5 times higher than the
Netherlands when the true difference is less than 10 percentage points.

• Dimensions

The standard X and Y axes are of course not the only way to plot data. Other
plotting dimensions also need to be used appropriately. These dimensions
include, but are not limited to: colour, shape, size, area, and transparency.

• Do not use more dimensions that are needed to be informative and

readable. Using more than 3 or 4 dimensions is often pushing the limits
of keeping a graph readable.

Page 33 of 64
 Presentation of Tables, Graphs and Statistics

• As with axes, differences between “values” of these dimensions should

be measurable and informative. For instance, detecting small differences
in transparency can be difficult. Therefore, using transparency is not
ideal for a continuous variable.
• Area is not necessarily a common dimension to use, but one type of
graph is quite dependent on it: a pie chart. While proportional bar charts
can convey the same message using area, these have the advantage of
an x/y axis showing the proportions, unlike pie charts.

While pie charts are common, they are rarely the most appropriate
choice. It is easy to find examples of using too many categories or
mismeasured segments, as seen below. It is also harder to read the area
of a circle than the height of a bar.

Therefore, pie charts are only useful with 4 or 5 categories and when the
differences between them are quite large. A bar chart will show small
differences much more clearly.

Page 34 of 64
 Presentation of Tables, Graphs and Statistics

• Orientation and ordering of your scales

As with tables, the orientation and ordering within your graph can make a big
difference to readability and can provide additional information.

Bar charts, box plots, violin plots etc. are often plotted with a continuous
variable along the y axis and categorical along the x axis. However, there are
advantages to switching this the other way around.

When dealing with large numbers of categories or long category labels,

flipping the x and y axis can improve the readability of your graph as you can
write the long labels without overlapping or having to turn them at an angle.

Also, consider the order of these categories the same way you would in a
table. Is there a logical order? If so, then stick to it. If not, consider putting
them in ascending or descending order by value. This makes the graph more
informative, as now the ranking will also be displayed.

Consider the next graph. The countries are in alphabetical order and the
rotated labels are hard to read. Comparing two bars of similar heights and
singling out the best and worst performers becomes challenging.

Page 35 of 64
 Presentation of Tables, Graphs and Statistics

This has been fixed by flipping the axes and ordering the countries by descending
order of the value on the x axis:

• Facets

A similar concept to scales is facets. This involves splitting a graph into

smaller, separate plots. These would be split by a categorical variable (maybe
two). You could think of facets as another scale - just one that splits up your
plotting space.

Using facets is a good way to avoid bars/lines etc. becoming too squashed
together or overlapping to the point where reading them becomes difficult.
They can also provide another layer of information to your results, as you can
compare across groups.

This means you could potentially use graph types which are limited in terms
of how many dimensions they offer. For instance, it is not advised to use
more than the x and y axis for a histogram, but you could split the histogram
over multiple smaller histograms and allow for comparability.

In the example on the next page, there are too many lines occupying the
same space, so it is difficult to pick out individual lines or make comparisons.
The graph is too visually noisy.

Page 36 of 64
 Presentation of Tables, Graphs and Statistics

We can instead split these lines by the type variable and plot five smaller,
individual line graphs which are easier to compare and single out.

Page 37 of 64
 Presentation of Tables, Graphs and Statistics

• Legends

See page 46

• Annotations

Placing annotations within the graph can be useful to help tell the story of
your data. They are especially useful for explaining unusual or interesting
data points such as outliers or spikes, as well as providing context or avoiding
misinterpretation.

• Using Colour

On any form of graph, colour is always an effective tool to differentiate

between different plotting elements. Often, we use colour to distinguish
between the levels of a categorical variable which may have also been
plotted along the x or y axis. However, if the variable has been plotted on an
axis already then adding colour is unnecessary and could just make your
graph visually noisy.

More guidance on colour will be provided in the section on accessibility.

• Using colour to highlight

o You can use colour to highlight key elements to aid interpretation.
This could be useful for showing quite specific messages about
your data.
For example, you may have multiple lines plotted on a line graph,
but you want to make a specific point about one or two of them
while the others are used to provide context or are useful but not
key to your message.
o Use a distinct colour in the foreground for the data you want to
draw attention too while shading the rest in a much more muted
colour, such as a pastel or shades of grey. This will mean the other
elements are still visible without them taking attention away from
your main point.
• Use colour with restraint. Using colour too much will make your graph
incredibly messy and create too much visual noise.
o Also, consider how your graph might be repurposed. Colour
should never be the only way meaning is added to your data. If
people print the document in black and white or are colour-blind,
will your message still be clear?

Page 38 of 64
 Presentation of Tables, Graphs and Statistics

o Use colour to mark helpful differences in your data. Different

colours suggest there is a meaning to those colours and if there is
not a useful meaning, then the audience wastes time trying to
figure out something that isn’t there.
o If you are creating a bar graph, it is easier to compare the bars
when they are the same colour. Colouring each bar differently is
unnecessary.

More Examples

Example 1: Poorly formatted and presented stacked bar chart

• No spacing between bars make comparisons difficult.

• No title.
• Uninformative axis labels.
• Uninformative legend.
• Red and green together is difficult to read and the colours are too saturated.
• Y axis is too large, making differences look smaller than they are.
• Harder to compare both within and across groups at the same time because
of use of counts and stacked bars.
• Poor spacing between axis and axis labels.

Page 39 of 64
 Presentation of Tables, Graphs and Statistics

Example 2: Improved alternative

• Informative title.
• Informative axis labels.
• Adequate spacing between plot and all textual elements.
• Colour-blind friendly colour palette.
• Reduced colour saturation.
• Black borders around bars to improve differentiation between bars and
colours.
• Use of percentages rather than counts, making it easier to compare both
between and within groups.
• Use of proportional stacked bars rather than stacks of counts. These are
easier to read when dealing with a categorical variable with two levels.
• More appropriate y axis scale.
• Increased spacing between bars to make them distinct.
• Included a source note in the bottom right of the plot.

Page 40 of 64
 Presentation of Tables, Graphs and Statistics

Example 3: Overplotting through limited number of unique values

Too many values crowd at the same point because there are only 4 possible unique
values along the x axis:

Example 4: Overplotting improved using jittered data points

This visualisation could also be improved by using alternative graph types such as a
boxplot, violin plot or a series of histograms.

Page 41 of 64
 Presentation of Tables, Graphs and Statistics

Examples 5 – 7: Making a line graph informative and readable

At first glance, we can still somewhat understand the patten and trend this graph is
showing us, but we are missing a lot of key information that allows this graph to be
understood out of context.

• No title
• Uninformative axis and legend labels
• Grey used as background colour
• Difficult to read text
• Lines are quite thin, making them harder to see clearly

Below, we have improved this example considerably by correcting these mistakes

according to the general guidance for graphs.

• Included a source
• Informative labelling and titles
• Better use of colour thanks to a white background
• Clearer axis lines
• Increased the thickness of the lines to stand out more

Page 42 of 64
 Presentation of Tables, Graphs and Statistics

This is a reasonably good example of a line graph but there are a couple more
refinements we can make.

• Directly labelling the lines has given us more plotting room.

An annotation has supplied additional contextual information that allows the
reader to see that this average was based off only 2 tv shows. This helps
prevent misinterpretation.

Page 43 of 64
 Presentation of Tables, Graphs and Statistics

Maps

Maps have become an increasingly common sight in our daily lives, being used to
show the variation in some statistic across the world or just a local area. Generally,
there are 2 main kinds of maps: reference maps and thematic maps. We are more
concerned with the latter, as reference maps are used to show an area’s geographic
extent and characteristics while thematic maps are used to display the geographic
distribution/variation in each variable.

In this sense, thematic maps are just another form of graph, albeit with a few
unique elements. They are useful when the geographic patterns of locations of your
data points are the most important part of the message to your readers. They can
be used for many different data types, including categorical and continuous data, as
well as aggregate measures (counts/percentages/densities) at some level of
geographical boundaries. (e.g., % unemployed within UK constituencies).

Before getting into the specifics of the maps themselves, it is worth considering a
few things about your data and how this could affect the map that you produce.

• Consider the level of geographic aggregation you use for your map. Of
course, this may depend on what geographic variables you have within your
data. However, if you do have multiple levels available, consider which is best
to show your results. Aggregation at higher levels may smooth out
differences at lower levels.
• Check you have enough data for your chosen level of aggregation. If some
areas have very few observations, you could be subject to extreme values
making interesting but inaccurate geographic patterns.

Page 44 of 64
 Presentation of Tables, Graphs and Statistics

Map Types

Choropleth Maps Dot Maps Proportional Symbol Maps

The most common type of thematic One of the few thematic map types which Symbols are placed in each geographic
maps. Data is summarised at a sensible does not involve geographic aggregation area of the map and are scaled in size
level of geographic aggregation. Values of your data. However, it will require very according to a variable of interest.
are then represented by shading in the low-level geographic information e.g., These tend to be hard to read as it can
different geographic areas in different geographic coordinates or a postcode. be difficult to compare relative sizes of
colours. Tends not to be used for totals Here each observation will be a point on the symbols. This can be helped by
as these will be affected by the size of a the map and are a good way to provide using a well-organised legend.
geographic area. an idea of the geographic density of your
data. These tend to be used when your
observations are locations or events to
show how they cluster in certain areas.

Page 45 of 64
 Presentation of Tables, Graphs and Statistics

Cartograms

These maps distort map geometry to show a specific message about the data. They can be quite effective if displayed alongside a
more standard choropleth map. However, being less common they can take some explanation when presented.

There are two distinct types:

Area cartograms Equal area cartograms

Map geometry is distorted to scale up areas All geographic areas are changed into shapes of the
which are high in the variable of interest and same size and positioned as best as possible to
scale down areas with low values. The size of an approximate their geographic relationships. Useful
area is proportional to its value. when the areas are very different in size, but area is
irrelevant to your message, as larger areas will
dominate the map. Appropriate labels may be needed
as it becomes difficult to identify specific areas.

Page 46 of 64
 Presentation of Tables, Graphs and Statistics

Examples:
Choropleth Map vs Cartogram - The examples here show how important it is to
use an appropriate style of map depending on the message you intend to show.

This is a US election
choropleth map showing
the results at the county
level. It offers a clear
message that most
counties were won by
Republican candidates.

However, this visual

message is misleading, as
election results are not
decided at the county
level, but rather at the
state level. Land does not
vote; people do.

The following cartogram offers a much more accurate picture of the US election.
The cartogram changes the states into coloured boxes which are sized
proportionally to the number of electoral colleges votes the states contribute. This
shows that the result is much closer than the county level map suggests and that
there not a significant lead in favour of the Republican Party.

Page 47 of 64
 Presentation of Tables, Graphs and Statistics

Using colour and categorising values

The advice presented in this section is tailored towards using colour scales in
choropleth maps or cartograms and would also be true of certain types of graphs.
Further advice on the use of colours will be provided in the accessibility section.

• If the variable of interest is on a continuous scale you should use different

shades of the same colour to fill in the geographic areas. Light shades should
signal low values and dark colours should represent high values.

• If you intend to map areas according to deviation from a given value such as the
average of the whole area, you should use a diverging colour scale. For instance,
using shades of red for below this value and blue for values above.

• If the variable is categorical and there is no natural order to them, use different
colours for each level.

• You should not use white except to represent 0, “no data” or the centre of a
diverging scheme.

Of course, using a colour scheme is only effective if you are breaking up the range
of values in a meaningful way. There are multiple ways you can achieve this, and the
choice will depend on your data. However, bear in mind that the choice of method
will greatly alter the appearance of the map. Generally, you should limit the scale to
avoid too many categories.

• Equal intervals – divide the data into equal class sizes. Useful with normally
distributed data to highlight a small number of areas. Not useful for skewed
data.
• Quantiles – assign an equal number of observations per class. Produces a
balanced map for non-uniformly distributed data. However, classes could
have vastly different ranges.
• Standard deviation – measure distance from the mean. Useful when the
data is normally distributed.

Page 48 of 64
 Presentation of Tables, Graphs and Statistics

• Manual adjustments – can be used to make more intuitive levels but may
make comparability across a series of maps difficult. May include some target
level to highlight areas below and above this value

Outliers should be highlighted as a unique category if they are important to the

messages from your data. If they are not important and are simply noise, they can
be hidden by widened ranges.

Map Elements

Title

As with graphs and tables, a map title should be concise but informative. It should
be larger than the rest of the text elements and centred at the top of the map. For a
map, the title will be required to include the “where” of the data as the geography is
central to the visualisation.

Legend

A legend is likely to be even more essential to a map than it was for a graph. When
creating a thematic map, we will almost always be using a colour axis or some other
symbology to tell a message about our data therefore a legend will almost always
be required. The following tips are also applicable to graph legends.

• If using proportional symbols, these should be the exact sizes they are on
the map.
• Preferably keep the symbol/colour to the left and the description on the right
• If any values are decimals below 1, then these should have a leading 0 and
not just a decimal point. (E.g., 0.52 rather than .52)
• Numbers in ranges should be separated using hyphens or the word ‘to’
o Avoid hyphens if said numbers are negative numbers. ‘-30 - -20’ is
difficult to read.
• If using different scales (e.g., both colour and proportional symbols), you can
organise them into groups separated by headings.

Scale

Maps are designed to be representative of a geographic area and we want our

visual to be as accurate to real geography as possible. Therefore, we use a
representation of the geographic area scaled down by some proportional
relationship. These are less important for thematic maps, but the scale of a map
should still be acknowledged.

Page 49 of 64
 Presentation of Tables, Graphs and Statistics

• You can use either a scale bar (a bar highlighting certain points showing the
real-life distance it represents.) or a textual representation (a sentence
describing the relationship between distance on the page and real-life
distances) or a representative fraction
o For a thematic map, a textual scale or representative faction are
arguably more appropriate and could be relegated to a footnote
• If using a scale bar.
o Avoid decimal numbers
o Make the bar long enough to be useful
o Round to easy to use, meaningful numbers
• The scale should be subtle and small, never drawing too much focus away
from the map body as it is there to provide context to the audience.

Inset Map/Locator Map

Inset or Locator maps are smaller secondary maps that are used to provide some
additional geography.

Inset maps are maps of related areas that are in some way separate from the main
geography, either because they do not border any of the other areas shown (such
as maps of the US separating out Alaska and Hawaii) or the areas on the inset map
are within the main map body but are too small to be adequately visualised on the
larger map body.

The latter is often seen on maps using UK constituencies to show the numerous
parts of London as there are many over a small geographic area and therefore
would be difficult to show on a map of the entire nation. They will use the same
colour scale and symbology as the primary map.

Locator maps are used to show the geographic context of the primary map. These
will tend to be off in the corner of the visual pane and will serve to place the main
geographic areas within a more familiar geographic context. For instance, if your
main map is areas of Bolivia, a locator map may be included to show where Bolivia
is within South America.

Using a locator map is useful when the area being shown is likely to be unfamiliar to
the map reader. It is useful to consider the potential geographic literacy of your
audience.

Page 50 of 64
 Presentation of Tables, Graphs and Statistics

Example
Bad Example

The example presented here makes some quite basic mistakes, mainly in the use of
colour and separation.

• A non-colour blind friendly colour scheme has been used to represent the
thematic data.
• Areas not being mapped thematically are a similar colour to bodies of water
so there is poor separation between the land and sea.
• Nation labels often overlap with lines representing national borders, making
them hard to read.
• The title is not very informative.
• The legend is untitled and uninformative.
• No scale included.
• Bold border lines draw unnecessary attention away from the main mapped
areas.

Page 51 of 64
 Presentation of Tables, Graphs and Statistics

Improved Example1

1
Different source of data but using the same variable is plotted. Bad example = ESS 2006, Improved example =
ESS 2018

Page 52 of 64
 Presentation of Tables, Graphs and Statistics

This improved version makes a considerable number of changes to create a much

neater and professional looking map.

• There is a bigger contrast of colours between the unmapped areas and

bodies of water.
• An improved colour scheme, which focuses solely on changes in luminance.
• A scale bar has been included.
• A much more explanatory title.
• A much neater and more informative legend.
• Annotations help provide additional interpretation of the material.
• Removed the nation labels to avoid clutter.

Accessibility

The following advice is adapted from advice on accessibility published by the

Government Statistical service (GSS), which can be found at the following web
address: https://gss.civilservice.gov.uk/policy-store/introduction-to-data-
visualisation/#section-8

While we may have specific target audiences in mind when we write up our results
and produce data visualisations, we should always aim to ensure inclusivity by
making our content accessible to those with certain impairments. The GSS specifies
impairments to vision, hearing, mobility and thinking & understanding as key areas
to consider. For data visualisations, this primarily concerns those with vision and
thinking impairments. These tips are designed to make your results as clear and
readable as possible in general.

Tables
• Use column headers which explain the content of the columns, including
measurement units where applicable.
• Include derived variables (e.g., totals) at the end of columns or rows.
• Try to use more rows than columns.
• Write out or clearly explain any acronyms.
• Use table footnotes/captions to provide extra important information that
cannot fit in the main body of the table.
• If you do not need to use exact numbers, consider rounding larger numbers.

Page 53 of 64
 Presentation of Tables, Graphs and Statistics

Graphs/Maps
• Write out or clearly explain any acronyms.
• Make sure there is a clear distinction between lines on a line graph.
• Do not use red and green together as it is difficult for colour-blind people to
distinguish between them.

Colour
Colour is one of the most useful tools for supplying extra information to maps and
graphs, and sometimes even tables. It can be used to clearly highlight patterns and
relationships that could be missed by a monochromatic visualisation. Adding colour
can make visualisations more effective, but this only works if viewers can tell which
colour is which. For instance, the graph presented here uses colours which are far
too similar.

Previous advice already presented on the use of colour in graphs and maps (using it
to highlight, use it sparingly etc.) are still relevant and contribute to ensuring
accessibility. The advice presented here tends to focus on the choice of colours
themselves.

• Ensure the colours are accessible.

o Colour blindness affects an individual’s ability to distinguish between
certain colours. It affects men more commonly than women.
o Most commonly, it affects the ability to distinguish reds and greens. Avoid
using these colours together. Less commonly, but to be considered, is
blue-yellow colour blindness. The two graphs below demonstrate how a
colour-blind person may see a graph using red and green. The colours
are much harder to distinguish.

Page 54 of 64
 Presentation of Tables, Graphs and Statistics

Original graph. Graph as a colour-blind person

might see it.

o Red-green colour palettes may also not be clear when printed in

greyscale.
o Blue palettes are a safe default starting point as they are colour blind safe
and are visible in grayscale, as shown in the picture below.

Original colours:

Colours converted
to greyscale:

• Choose colours carefully.

o Consider cultural context. Colours often have some inherent culturally
defined associations. For instance, using colours people associate with
familiar concepts can improve the quality and speed of information
processing, such as using blue for water on a map.

Page 55 of 64
 Presentation of Tables, Graphs and Statistics

• Understand the digital colour palette.

o Colours are represented using several common schemes. The most
useful of these considers hue, saturation, and luminance. This scheme
allows us to intuitively define unique colours.
o Hue – Hues are colours. They do not have a natural order and therefore
users cannot assign a logical order to them. Small changes in hue are
easy to detect although colour blindness can affect how well people can
detect these differences.

o Saturation – This is the intensity of a colour, ranging from grey/white (no

saturation) to rich, vibrant, almost glowing colour. Saturation is perceived
on a continuous scale, although it is difficult to detect small changes. High
saturation can also cause issues for those with certain visual/light
sensitivity problems.

o Luminance – the brightness of colour. Also perceived as continuous

ordered scale from dark to light. This natural order can help us optimize
colour schemes for maximum distinction and differentiation.
▪ Changes are easier to detect
▪ It is easier to distinguish between bars even if luminance is the
only difference.
▪ Changes in luminance need to be larger if creating a line graph as
the white space between the lines makes it harder to quickly
compare.

Page 56 of 64
 Presentation of Tables, Graphs and Statistics

• Never use an image as a background. This looks messy and cluttered and can
make it difficult to read the graph.

• Know how to use colour effectively

o Alternate colours – consider alternating dark and light for categorical data
to improve clarity and distinction.

o Use borders – adding thin borders to the edges of bars can enhance
clarity/separation.
o Avoid overuse of saturated colours.
▪ Mid to low saturations are preferred
▪ Only use bold saturated colour to draw attention to a specific piece
of information or hard to see, small elements.

Page 57 of 64
 Presentation of Tables, Graphs and Statistics

▪ Bold, saturated colours can have visual side-effects. They may

appear to glow for many users, can generate after-images and can
affect how other colours appear.
▪ The first graph below uses very saturated and bright colours which
are quite straining to the eye. The second keeps the same hue but
reduces the saturation and luminance and adds in some border
lines for better separation.

o Use colour logically and consistently. If using sequences of colours,

ensure that they progress in a logical manner that the user would expect,
such as increasing luminance like in the map on page 50.
▪ If creating multiple graphs, use the same colour to mean the same
thing. Changing what they mean can confuse the user.
o Use a white background.
▪ Most palettes are designed to appear on top of a white
background. It provides a helpful reference for the colour scale
being used.

Page 58 of 64
 Presentation of Tables, Graphs and Statistics

Advice on Software

This guide has avoided specific guidance for creating visuals in any particular
software, as preferences will vary between readers and over time.

However, we will set out some general advice for you to consider based on the
guidance offered throughout this resource. We have repeatedly stressed the
importance of design, readability, and consistency. Therefore, there are a few
characteristics your software should preferably adhere to.

Your chosen software should offer the ability to code, program, or edit all elements
of the graph, so you can quickly draw, modify, or share the visualisation. Preferably,
it should also be open source/free so that you can learn from others and keep your
costs down.

As such, we generally recommend using R/RStudio and particularly the package

ggplot2, which is designed for effective creation of graphs inspired by good practice
and principles. For creating tables and maps, other software may be better suited
for your purposes, although R is also capable of producing these through well-
documented packages such as FlexTable or Leaflet (for maps).

Page 59 of 64
 Presentation of Tables, Graphs and Statistics

Checklist

For all your data visualisations, consider these key points as a form of a checklist,
making sure to mark off all points which are relevant to the work you are producing.

General Guidance
□ Effectiveness
o Are you using the right kind of visualisation?

□ Readability
o Are your titles and labels clear and informative?
o Is it suitable for your target audience?

□ Tidiness
o Have you avoided clutter?
o Are the elements well spaced out and positioned?

□ Accessibility
o Have you explained all abbreviations?
o Have you used colour blind friendly schemes?

□ Consistency
o Have you been consistent between your tables and graphs?
▪ Kept categories in same order?
▪ Kept the meaning of colours the same?

□ Informative
o Does your visualisation make sense out of context?
o Have you included the source?

Tables
□ Have you kept the level of precision consistent?

□ Have you kept the number of decimals to the minimum required for your
results to be informative?

□ If dealing with very big or small numbers have you considered

standardisation or scientific notation?

□ Remember it is easier to compare numbers in columns than in rows

□ Is the order of your rows?

Page 60 of 64
 Presentation of Tables, Graphs and Statistics

o Logical?
o Ascending/Descending?
o Have none/other categories been put at the bottom of the table?

□ Have you only used border sparingly?

o Avoided putting them within the main body of the table?
o Only used horizontal lines?

□ Been consistent and professional with font

□ Grouped columns or rows where possible?

Graphs
□ Are you using the right kind of graph?

□ Are all elements well-spaced out and not cluttered?

□ Have you avoided “junk” features?

□ Have you avoided overplotting?

□ Have you avoided plotting too much information?

□ Have you used and appropriate number of dimensions/variables?

□ Are your axes well labelled and appropriate for the type of graph you are
making?

□ Is there a clear distinction between the different levels of other dimensions

you may be using?

□ Would annotations be additionally informative to your viewers?

□ Have you used colour effectively?

o Consider:
▪ Highlighting.
▪ Have you used restraint?
▪ Is it colour blind friendly?
▪ Is it easy on the eyes (not overly saturated)?

Maps
□ Are you using the right kind of map?

□ Are the colour scale and categories of values you are using appropriate and
well labelled?

Page 61 of 64
 Presentation of Tables, Graphs and Statistics

□ Is your legend accurate and informative?

□ Have you included a well formatted scale bar?

□ Would interpretation be aided using a locator or inset map?

o Are these appropriately positioned and understandable?

□ Do the most important elements (the thematic symbols) stand out the most?

Page 62 of 64
 Presentation of Tables, Graphs and Statistics

Other Resources

General

GSS – Introduction to Data visualisation

https://gss.civilservice.gov.uk/policy-store/introduction-to-data-
visualisation/#section-7

Informative Presentation of Tables, Graphs and Statistics

https://stats4sd.org/resources/412

Data visualisation examples

https://stats4sd.org/resources/59

Tips on effective use of tables and figures in research papers

https://www.editage.com/insights/tips-on-effective-use-of-tables-and-figures-in-
research-papers

Tufte Principles
https://sites.google.com/site/tufteondesign/home/six-fundamental-principles-of-
design

Tables

Exporting Tables from R using Flextable, Kable and gt

https://stats4sd.org/resources/506

Preparing tables for research papers

https://www.manuscriptedit.com/scholar-hangout/preparing-tables-research-
papers/

Formatting Tables in MS word

http://www.docs.is.ed.ac.uk/skills/documents/3575/3575.pdf

Graphs

R Graphics Cookbook
https://r-graphics.org/

Financial Times – Visual Vocabulary

https://github.com/ft-interactive/chart-doctor/blob/master/visual-vocabulary/Visual-
vocabulary.pdf

Page 63 of 64
 Presentation of Tables, Graphs and Statistics

Maps

ESRI lecture – Map Elements and Design Principles

https://www.arcgis.com/apps/Cascade/index.html?appid=e739c503a1f04d3883983
4a0fe4ca6d4

ESRI – Make Maps People Want to Look At

https://www.esri.com/news/arcuser/0112/make-maps-people-want-to-look-
at.html#:~:text=Five%20of%20the%20main%20design,map%20and%20on%20the%
20page

Carto – 6 Design Principles for Making Maps on the Web

https://carto.com/blog/six-design-principles-making-maps-on-the-web/

GISLounge – Ten Things to Consider When Making a Map

https://www.gislounge.com/ten-things-to-consider-when-making-a-map/

GSS – Effective Maps in official statistics

https://gss.civilservice.gov.uk/wp-
content/uploads/2020/06/Effective_Maps_temporary_May_2020.pdf

Accessibility

ONS – Web accessibility

https://style.ons.gov.uk/writing-for-the-web/web-accessibility/introduction-3/

Tableau – Colour Blindness

https://www.tableau.com/about/blog/2016/4/examining-data-viz-rules-dont-use-
red-green-together-53463

Sources:

USA election maps:

https://theconversation.com/how-to-read-u-s-election-maps-as-votes-are-being-
counted-149251

Misleading Gun death graph

https://www.livescience.com/45083-misleading-gun-death-chart.html

Find more atstats4sd.org/resources Page 64 of 64