Please complete the Student Voice section on NOW.
Where We Are Starting From
Weeks 2 To 5 Build One Workflow
Week 2: Principles of data visualisationWeek 3: Grammar of graphics; aesthetics and attributesWeek 4: Major visualisation tools- Week 5: Customising visualisations (scales, themes, and labels)
What Week 5 Adds
By the end of this session, you should be able to:
- Improve
ggplot2visualisations using labels, legends, colour, and simple themes. - Choose aesthetics and scales that fit the variable type and comparison.
- Decide which customisation improves clarity, and which is only decoration.
- Start a formative assessment RMarkdown using either the provided dataset or your own dataset.
Download the Week 5 files from NOW. Unzip the folder, then move the unzipped Week 5 folder into your existing RStudio Project folder.
Today: Practical Scope
The session has two practical blocks.
| Block | Main work | Files |
|---|---|---|
| 1 | Make plots easier to understand with labels, legends, colour, and aesthetics. |
01_labels_and_legends.Rmd02_colours_and_aesthetics.Rmd selected parts
|
| 2 | Start a formative assessment figure and decide what the plot should show. |
04_formative_practice_provided_data.Rmdor 05_formative_planning_own_data.Rmd
|
03_themes.Rmd is optional/homework/reference material for after the session.
The priority today is not to try every customisation option. It is to make better decisions about which changes help the viewer.
Discussion: What Helps You Read This Chart?
What do you notice first? Which labels or colours would you change?
Source: Office for National Statistics, Figure 3 (2025). Decile 1 = most deprived areas; 10 = least deprived.
Variable Type Also Guides Aesthetics
In Week 4, variable type helped us choose geoms. In Week 5, it helps us choose aesthetics and scales.
| Variable type | Good aesthetic choices | Be careful with |
|---|---|---|
| categorical groups | colour, fill, shape, linetype, facets | too many colours or shapes |
| ordered categories | ordered colour, position, facets | unordered palettes |
| continuous values | position, colour gradient, size | size can be hard to compare |
| binary variables | colour, shape, linetype | redundant legends |
Aesthetic mappings should make the important comparison easier, not just make the plot more colourful.
Customisation Serves The Comparison
Customisation should help the viewer answer the question.
| Customisation | Helps the viewer… | Ask yourself… |
|---|---|---|
| labels | know what the variables, units, and groups mean | could this plot stand alone in a report? |
| legends and facets | understand which observations belong to which group | are the group names meaningful? |
| colour and aesthetics | compare categories, ordered values, or continuous values | does the aesthetic fit the variable type? |
| scales | read values fairly and accurately | could the scale mislead the comparison? |
| themes | remove unnecessary distraction | did this remove clutter or add decoration? |
| export choices | make the figure readable at final size | are text, points, and aspect ratio still clear? |
Week 5 is about defensible visualisation choices, not polishing plots for its own sake.
Customisation Applies Visualisation Principles
Tufte’s principles are not just about making sparse plots. They ask whether every visible element helps the viewer understand the data.
| Principle | What customisation can do | Example Week 5 choice |
|---|---|---|
| increase data-ink | make the data easier to see than the decoration | lighten or remove heavy grid lines |
| reduce chartjunk | remove marks that do not support the comparison | avoid decorative themes and unnecessary colours |
| support comparison | keep related plots comparable | use consistent scales, facets, and labels |
| guide interpretation | help viewers notice the intended pattern | use titles, subtitles, captions, and direct labels |
| avoid distortion | make readings fair and transparent | choose appropriate axes, scales, and grouping aesthetics |
Customising a plot is an argument about what matters in the data.
Design Choices Should Reduce Distraction
Revisiting Tufte’s Principles (Tufte, 2001):
- Data-ink ratio: make the data marks more prominent than the supporting marks.
- Chartjunk: remove decoration that does not help the comparison.
- Small multiples: use consistent scales and layouts so groups can be compared fairly.
- Narrative graphics: use text, annotations, and labels to guide interpretation.
The point is not that every plot should be minimal. The point is that every customisation should earn its place.
Tufte Ideas Become ggplot2 Decisions
In ggplot2, these principles become specific choices we can make and justify.
| Tufte principle | ggplot2 decision |
|---|---|
| increase data-ink | use quiet themes, lighter grids, and visible data marks |
| reduce chartjunk | avoid decorative palettes, heavy backgrounds, and unnecessary legends |
| support comparison | use facets, consistent scales, and aligned axes |
| guide interpretation | write meaningful titles, axis labels, legend labels, and captions |
| avoid distortion | choose sensible scales, transparency, jittering, or grouping aesthetics |
The grammar of graphics helps because each layer can be changed deliberately: data, mapping, geom, scale, coordinate system, facet, and theme.
Labels Make Plots Stand Alone
titlesubtitlecaptiontagxycolour,shape,linetype,fill
ggplot(d_spellname, aes(y = rt,
x = dur,
colour = modality)) +
geom_point() +
labs()
Titles Should Name The Comparison
titlesubtitlecaptiontagxycolour,shape,linetype,fill
ggplot(d_spellname, aes(y = rt,
x = dur,
colour = modality)) +
geom_point() +
labs(title =
"Reaction time and duration")
Subtitles Add Context
titlesubtitlecaptiontagxycolour,shape,linetype,fill
ggplot(d_spellname, aes(y = rt,
x = dur,
colour = modality)) +
geom_point() +
labs(title =
"Reaction time and duration",
subtitle = "Participant-level averages")
Captions Tell Readers Where Data Came From
titlesubtitlecaptiontagxycolour,shape,linetype,fill
ggplot(d_spellname, aes(y = rt,
x = dur,
colour = modality)) +
geom_point() +
labs(caption = "Data: picture naming norms")
Tags Can Label Panels
titlesubtitlecaptiontagxycolour,shape,linetype,fill
ggplot(d_spellname, aes(y = rt,
x = dur,
colour = modality)) +
geom_point() +
labs(tag = "A")
Axis Labels Need Units
titlesubtitlecaptiontagxycolour,shape,linetype,fill
ggplot(d_spellname, aes(y = rt,
x = dur,
colour = modality)) +
geom_point() +
labs(x = "Production duration (ms)",
y = "Reaction time (ms)")
Legend Titles Name The Grouping
titlesubtitlecaptiontagxycolour,shape,linetype,fill
ggplot(d_spellname, aes(y = rt,
x = dur,
colour = modality)) +
geom_point() +
labs(colour = "Response mode")
Exercise 1: Changing Text
Open 01_labels_and_legends.Rmd.
Complete Tasks 1-3 (about 12 minutes): choose axis labels, improve the legend, and write a title.
Keep one labelled figure and one sentence explaining your choice.
Stop at “In-class stopping point”. The remaining label examples are optional.
Colour: Hue Versus Brightness
Colour can represent different kinds of information.
| Colour property | Best for | Example |
|---|---|---|
| Hue | categories | speech vs writing |
| Brightness / lightness | ordered or continuous values | low to high vocabulary |
| Diverging colour | values around a meaningful midpoint | below vs above average |
Use different hues for different kinds of things. Use light-to-dark progressions for more/less.
Colour Scales Change Values And Labels
scale_colour_discretescale_colour_continuousscale_colour_manual- or any other aesthetic instead of
colour
ggplot(d_spellname, aes(y = rt,
x = dur,
colour = modality)) +
geom_point() +
scale_colour_discrete(
labels = c("Spoken output",
"Written output"))
Existing Palette Functions Are The Easiest Start
- Use a palette function first
ggthemes::scale_colour_colorblind()- Then customise manually only when needed
ggplot(d_spellname, aes(y = rt,
x = dur,
colour = modality)) +
geom_point() +
scale_colour_colorblind(
labels = c("Spoken output",
"Written output"))
Manual Colours Should Support Comparison
- Change colour values manually
- Colour names: link
ggthemes
ggplot(d_spellname, aes(y = rt,
x = dur,
colour = modality)) +
geom_point() +
scale_colour_manual(
labels = c("Spoken output",
"Written output"),
values = c("#0072B2", "#D55E00"))
Store Palette Values When You Need Manual Control
- Start from a tested palette
- Store hex codes if you need to reuse them
scales::show_col()previews colours
# RGB codes of "colorblind" function
mycolours <- c("#000000", "#E69F00",
"#56B4E9", "#009E73",
"#F0E442", "#0072B2",
"#D55E00", "#CC79A7")# RGB codes of "colorblind" function scales::show_col(colorblind_pal()(8))
Legend Labels Should Be Human-Readable
- Replace data codes with readable group names
labelschanges the legend wordingvaluescontrols the corresponding colours
ggplot(d_spellname, aes(y = rt,
x = dur,
colour = modality)) +
geom_point() +
scale_colour_manual(
labels = c("Spoken output",
"Written output"),
values = mycolours[1:2])
Discussion: Colour, Shape, Or Facets?
Each version shows the same relationship. Which would you use, and what trade-off does it make?
When does another aesthetic help, and when does it become clutter?
Exercise 2: Colour And Aesthetic Decisions
Open 02_colours_and_aesthetics.Rmd.
Complete Tasks 1-2 (about 12 minutes): choose a palette and compare overlapping age distributions.
Record which version you would use and what the viewer can compare more easily.
Stop at “In-class stopping point”. Manual palette exploration is optional.
Facet Strips Are Labels Too
ggplot(d_spellname, aes(y = rt,
x = dur)) +
geom_point() +
facet_grid(~modality)
Facet Labels Can Come From The Data
ggplot(d_spellname, aes(y = rt,
x = dur)) +
geom_point() +
facet_grid(~modality,
labeller = label_both)
Recoding Changes The Data Labels
d_spellname <- mutate(d_spellname,
modality = recode(modality,
speech = "Spoken output",
writing = "Written output"))
ggplot(d_spellname, aes(x = dur, y = rt)) +
geom_point() +
facet_grid(~modality)
Strip Labels Should Match The Story
ggplot(d_spellname, aes(y = rt,
x = dur)) +
geom_point() +
facet_grid(~modality, labeller = label_both)
Themes Control Non-Data Ink
Start with a complete theme, then change individual elements with theme().
| Complete theme | Starting appearance |
|---|---|
theme_grey() |
grey panel and white grid lines |
theme_bw() |
white panel, grid lines, and a border |
theme_classic() |
axis lines without grid lines |
theme_minimal() |
grid lines without a panel background |
theme_light() |
light grid lines and a border |
theme_dark() |
dark panel and lighter grid lines |
ggthemes adds starting points such as theme_clean() and theme_tufte().
Choose a starting point that makes the comparison easy to read.
Start From The ggplot2 Default
ggplot(d_spellname, aes(y = rt,
x = dur)) +
geom_point() +
facet_grid(~modality) +
theme_grey(base_size = 11)
theme_minimal(): No Panel Background
ggplot(d_spellname, aes(y = rt,
x = dur)) +
geom_point() +
facet_grid(~modality) +
theme_minimal(base_size = 14)
theme_light(): Light Grid And Border
ggplot(d_spellname, aes(y = rt,
x = dur)) +
geom_point() +
facet_grid(~modality) +
theme_light(base_size = 14)
theme_dark(): A Dark Panel Background
ggplot(d_spellname, aes(y = rt,
x = dur)) +
geom_point() +
facet_grid(~modality) +
theme_dark(base_size = 14)
Theme Packages Add Starting Points
ggplot(d_spellname, aes(y = rt,
x = dur)) +
geom_point() +
facet_grid(~modality) +
theme_clean()
Choose Themes For The Comparison
axislegendpanelplotstrip
ggplot(d_spellname, aes(y = rt,
x = dur)) +
geom_point() +
theme()
Axis Text Is Part Of Readability
axis.textaxis.text.xaxis.text.y
axis.titleaxis.title.xaxis.title.y
ggplot(d_spellname, aes(y = rt,
x = dur)) +
geom_point() +
theme(axis.text = element_text(
face = "italic"))
Bold Axis Titles Can Add Emphasis
axis.textaxis.text.xaxis.text.y
axis.titleaxis.title.xaxis.title.y
ggplot(d_spellname, aes(y = rt,
x = dur)) +
geom_point() +
theme(axis.title = element_text(
face = "bold"))
Axis Titles And Tick Labels Are Separate
axis.textaxis.text.xaxis.text.y
axis.titleaxis.title.xaxis.title.y
ggplot(d_spellname, aes(y = rt,
x = dur)) +
geom_point() +
theme(axis.title.y = element_text(
face = "bold"))
Legends Compete For Plot Space
legend.backgroundlegend.marginlegend.spacinglegend.keylegend.textlegend.titlelegend.positionlegend.orientationlegend.justificationlegend.box
ggplot(d_spellname, aes(y = rt,
x = dur,
colour = modality)) +
geom_point() +
theme()
Move Legends When Space Is Tight
legend.backgroundlegend.marginlegend.spacinglegend.keylegend.textlegend.titlelegend.positionlegend.orientationlegend.justificationlegend.box
ggplot(d_spellname, aes(y = rt,
x = dur,
colour = modality)) +
geom_point() +
theme(legend.position = "top")
Align A Top Legend To The Right
legend.backgroundlegend.marginlegend.spacinglegend.keylegend.textlegend.titlelegend.positionlegend.orientationlegend.justificationlegend.box
ggplot(d_spellname, aes(y = rt,
x = dur,
colour = modality)) +
geom_point() +
theme(legend.position = "top",
legend.justification = "right")
An Inside Legend Can Cover Data
legend.backgroundlegend.marginlegend.spacinglegend.keylegend.textlegend.titlelegend.positionlegend.orientationlegend.justificationlegend.box
ggplot(d_spellname, aes(y = rt,
x = dur,
colour = modality)) +
geom_point() +
theme(legend.position = c(.15, .8))
Grid Lines Can Help Or Distract
panel.backgroundpanel.borderpanel.spacingpanel.gridpanel.grid.majorpanel.grid.minor
ggplot(d_spellname, aes(y = rt,
x = dur)) +
geom_point() +
theme()
Remove The Panel Background
panel.backgroundpanel.borderpanel.spacingpanel.gridpanel.grid.majorpanel.grid.minor
ggplot(d_spellname, aes(y = rt,
x = dur)) +
geom_point() +
theme(panel.background =
element_blank())
Plot Properties Control The Whole Figure
plot.backgroundplot.marginplot.titleplot.subtitleplot.captionplot.tag
ggplot(d_spellname, aes(y = rt,
x = dur)) +
geom_point() +
theme()
A Coloured Plot Background Adds Non-Data Ink
plot.backgroundplot.marginplot.titleplot.subtitleplot.captionplot.tag
ggplot(d_spellname, aes(y = rt,
x = dur)) +
geom_point() +
theme(plot.background = element_rect(
fill = "pink"))
Plot Margins Affect Readability
plot.backgroundplot.marginplot.titleplot.subtitleplot.captionplot.tag
ggplot(d_spellname, aes(y = rt,
x = dur)) +
geom_point() +
theme(plot.background = element_rect(
fill = "pink"),
plot.margin = unit(c(2, 2, 2, 2),
"cm"))
Title Colour Can Help Or Hinder Readability
plot.backgroundplot.marginplot.titleplot.subtitleplot.captionplot.tag
ggplot(d_spellname, aes(y = rt,
x = dur)) +
geom_point() +
labs(title = "I'm a title") +
theme(plot.title = element_text(
colour = "pink"))
Caption Text Can Be Styled Separately
plot.backgroundplot.marginplot.titleplot.subtitleplot.captionplot.tag
ggplot(d_spellname, aes(y = rt,
x = dur)) +
geom_point() +
labs(caption = "I'm a caption") +
theme(plot.caption = element_text(
face = "italic"))
Facet Strips Guide Small Multiples
strip.backgroundstrip.placementstrip.text
ggplot(d_spellname, aes(y = rt,
x = dur)) +
geom_point() +
facet_grid(~modality,
labeller = label_both) +
theme()
Strip Backgrounds Can Add Clutter
strip.backgroundstrip.placementstrip.text
ggplot(d_spellname, aes(y = rt,
x = dur)) +
geom_point() +
facet_grid(~modality,
labeller = label_both) +
theme(strip.background =
element_blank())
Strip Background Colour Changes The Emphasis
strip.backgroundstrip.placementstrip.text
ggplot(d_spellname, aes(y = rt,
x = dur)) +
geom_point() +
facet_grid(~modality,
labeller = label_both) +
theme(strip.background = element_rect(
fill = "forestgreen"))
Strip Text Needs To Stay Readable
strip.backgroundstrip.placementstrip.text
ggplot(d_spellname, aes(y = rt,
x = dur)) +
geom_point() +
facet_grid(~modality,
labeller = label_both) +
theme(strip.background = element_rect(fill = "forestgreen"),
strip.text = element_text(colour = "white", hjust = 0))
Combine Theme Changes Deliberately
strip.backgroundstrip.placementstrip.text
ggplot(d_spellname, aes(y = rt,
x = dur)) +
geom_point() +
facet_grid(~modality,
labeller = label_both)+
theme(strip.background = element_rect(fill = "forestgreen"),
strip.text = element_text(colour = "white", hjust = 0,
face = "bold", size = 16, angle = 180))
Optional/Homework: Themes
03_themes.Rmd is optional/homework/reference material.
If you have time, use it to compare themes and choose a quiet starting point for an assessment-style figure. The aim is not to design a personal style from scratch during class.
Useful Resources For Data Visualisation
- Choosing the right visualisation: data-to-viz.com
- Rick Scavetta’s course: datacamp.com
ggplot2CheatSheet- Most important functions: ggplot2.tidyverse.org
Recommended Reading
For support with this week’s material, use:
- Andrews (2021), Chapter 4: data visualisation in R. Online chapter
- The
ggplot2online book sections on scales, labels, themes, and saving plots as optional extra support. Online book
Start The Formative Figure
This is the main second-block task. Use one of the two templates in the Week 5 files:
04_formative_practice_provided_data.Rmdif you want to practise with the provided Blomkvist dataset.05_formative_planning_own_data.Rmdif you already have a candidate dataset.
Start by deciding what your first figure should help the viewer compare.
Use the next 25 minutes to:
- Choose one question and draft one figure.
- Give the document a title about your data or question.
- Give the figure section a heading about the comparison.
- Add a caption and explain one design choice in 2-3 sentences.
You can run chunks while developing the figure.
Formative Assessment: The Full Brief
Select a behavioural dataset, such as reaction times, survey responses, or decision-making data.
- Create 3-4 visualisations using different techniques (e.g., boxplot, scatterplot, density plot).
- Include brief commentary on:
- What each visualisation highlights about the data?
- Why each visualisation was chosen?
- How each visualisation adheres to principles of effective visualisation (e.g., Tufte’s guidelines)?
- What each visualisation may obscure about the data.
- Deadline: Friday 30 October 2026, before 2 pm
- Length: max 1,000 words
Feedback focus:
- Clarity and effectiveness of visualisations.
- Code quality and reproducibility.
- Interpretation and communication of behavioural insights.
Explain The Choice In Your Own Figure
Choose one or two decisions in your figure. Explain how they help answer your question and what the figure still hides.
| Decision | Options you could compare |
|---|---|
| Showing variation | raw points, a summary, or both |
| Distinguishing groups | colour, colour and shape, or separate facets |
| Showing overlap | opaque fill, transparent fill, or outlines |
| Making text useful | variable names or descriptive labels with units; a question or finding as the title |
| Keeping useful structure | keep, lighten, or remove grid lines; choose a starting theme |
Example: “I used transparent fills so the overlapping age distributions remain visible, but the curves do not show how many participants are in each group.”
References
Andrews, M. (2021). Doing data science in R: An introduction for Social Scientists. SAGE Publications Ltd.
Tufte, E. R. (2001). The visual display of quantitative information (2nd ed.). Graphics Press.