Advanced Statistical Programming using R

Week 11: Statistical Communication & Visualisation

2026-06-25

Announcements

Upcoming deadlines

• Progress Update due Jun 29
  • Submit links to rendered website, project plan, and GitHub repo via Moodle
  • We will provide group-specific feedback via GitHub issues
• Exam registration closes Jun 29
  • see Moodle announcement

This week

• Group Project
• Communicating with Data
  • Tables with {gt}
  • Visualisations with {ggplot2}

Group Project

Key dates

Milestone	Date	Submission
Draft Proposal	Jun 8	✓ general feedback in class today
Progress Update	29 Jun	URL to rendered site, updated Project Proposal
Group reflection — in-class discussion	W14 practical	N/A
Final submission + group-reflection.qmd	Due 22 Jul	URL to final website/webpage, group reflection doc
Individual contribution statement	Due 23 Jul	PDF via Moodle
Oral exam	29 Jul	—

From Proposals to Project Plans

• Submitted proposals are Draft Project Plans:
  • datasets & IDA: what data are you exploring?
  • research questions: what are you trying to investigate or learn?
  • analysis methods: how will you investigate your questions?
  • data preparation: what preparation is required to interpret your data and conduct your planned analyses?
  • final outputs: what formats will you use to document and present your analysis process and findings?

Progress Update

• After last week, you should have some descriptive, explanatory or predictive findings.
• This week, we switch towards communicating your data analysis and findings.

Last Week

Syllabus

Part 1: Statistical Programming Foundations

• W02: Scripts, Functions & Refactoring
• W03: Debugging
• W04: Version Control & Remotes
• W05: Quarto Websites & Collaborative Coding
• W06: R Packages

Part 2: Working with Real World Data

• W07: Initial Data Analysis & Data Cleaning
• W09: Reproducibility, Open Data & renv
• W10: Analysis & Inference
• W11: Statistical Communication & Visualisation

Part 3: Advanced Topics & Summary

• W12: Interactive Data Storytelling
• W13: Multi-lingual Analysis (Python/R)
• W14: Review & Outro

Learning from data

• Three goals: descriptive (what is the state of things?), explanatory (is there a real relationship?), predictive (does this generalise?)
• Exploratory vs confirmatory is orthogonal to the goal — any analysis type can be either; the distinction is whether the question was fixed before looking
• Inference tools differ by goal: CIs for descriptive, hypothesis tests + effect sizes for explanatory, held-out evaluation for predictive

Planning for inference

• A data analysis plan covers the full project; an inference plan is the component where you generalise beyond your sample
• Running tests after exploring is normal — label those findings as exploratory/candidate and treat them as hypotheses for future work
• Each look at the data spends some of the evidentiary budget — standard tools don’t adjust for this automatically
• Write it down: even a paragraph before you start is enough to separate what you planned from what you discovered

{infer}

• Workflow: specify() → assume() → calculate() → visualize() / get_p_value()
• assume("t" / "z" / "Chisq" / "F") uses the same theoretical distributions as Stat 1/2
• The null plot shows you where your test statistic sits in the null distribution
• Extension: generate() simulates the null without assuming a distribution

Communicating with Data

Why data communication skills matter

So far you’ve been learning skills and tools for examining and learning from data. However, it is also important to be able to clearly and effectively communicate your findings and how you reached them. Effective communication supports:

• Clarity — well-communicated insight is more useful than a technically correct but unreadable analysis
• Persuasion — analysis only creates impact if it reaches and convinces decision-makers
• Detecting misinformation — understanding how data can mislead makes you a better producer and consumer of analysis

Communication tools for statistical information

We use tables and plots to show:

• descriptive statistics
• patterns and relationships between variables
• regression results
• model uncertainty and diagnostics

Choosing between them depends on:

• what you want to show
• whether you want to display exact values or communicate a message
• who your audience is

Descriptive summary statistics

A numerical summary that gives a “feel” of what the data contains should convey:

• variables in the data and number of observations
• missing values (if any)
• distribution — five-number summaries, or counts/percentages for categorical variables
• correlations between variables

Correlation Tables

library(corrr)
library(gt)
library(palmerpenguins)
library(dplyr)

penguins |>
  select(bill_length_mm, bill_depth_mm, flipper_length_mm, body_mass_g) |>
  correlate(quiet = TRUE) |>
  shave() |>
  gt() |>
  cols_label(term = "") |>
  fmt_number(where(is.numeric), decimals = 2) |>
  sub_missing(missing_text = "") |>
  tab_header(title = "Palmer Penguins — correlation matrix")

Correlation Tables

Palmer Penguins — correlation matrix
	bill_length_mm	bill_depth_mm	flipper_length_mm	body_mass_g
bill_length_mm
bill_depth_mm	−0.24
flipper_length_mm	0.66	−0.58
body_mass_g	0.60	−0.47	0.87

Correlation Plots

library(corrr)
library(palmerpenguins)
library(dplyr)

penguins |>
  select(bill_length_mm, bill_depth_mm, flipper_length_mm, body_mass_g) |>
  correlate(quiet = TRUE) |>
  rplot() +
  theme_minimal() +
  scale_x_discrete(position = "top") +
  labs(title = "Palmer Penguins — correlation plot")

Correlation Plots

Cross-Tabulations

Cross-tabulations (contingency tables) show the frequency distribution of two or more categorical variables simultaneously — how many observations fall into each combination of categories.

library(gt)
library(palmerpenguins)
library(dplyr)
library(tidyr)

penguins |>
  count(species, island) |>
  pivot_wider(names_from = island, values_from = n, values_fill = 0) |>
  gt() |>
  cols_label(species = "Species") |>
  tab_spanner(label = "Island", columns = -species) |>
  tab_header(title = "Palmer Penguins by species and island")

Cross-Tabulations

Palmer Penguins by species and island
Species	Island
	Biscoe	Dream	Torgersen
Adelie	44	56	52
Chinstrap	0	68	0
Gentoo	124	0	0

Model Results & Diagnostics

The key numerical characteristics of a fitted model typically include:

• parameter estimates — coefficients, odds ratios, hazard ratios
• uncertainty — standard errors, confidence/credible intervals
• model fit — R², AIC, BIC, RMSE, or other diagnostics appropriate to the model type

Model Results Table

library(broom)
library(gt)
library(palmerpenguins)

lm(body_mass_g ~ bill_length_mm + species, data = penguins) |>
  tidy(conf.int = TRUE) |>
  gt() |>
  fmt_number(where(is.numeric), decimals = 3) |>
  cols_label(
    term      = "Term",
    estimate  = "Estimate",
    std.error = html("Std.<br>Error"),
    statistic = md("*t*"),
    p.value   = md("*p*"),
    conf.low  = html("95% CI<br>lower"),
    conf.high = html("95% CI<br>upper")
  ) |>
  tab_header(
    title    = "Linear regression: penguin body mass",
    subtitle = "Predictors: bill length and species"
  )

Model Results Table

Linear regression: penguin body mass
Predictors: bill length and species
Term	Estimate	Std. Error	t	p	95% CI lower	95% CI upper
(Intercept)	153.740	268.901	0.572	0.568	−375.191	682.670
bill_length_mm	91.436	6.887	13.276	0.000	77.889	104.983
speciesChinstrap	−885.812	88.250	−10.038	0.000	−1,059.401	−712.223
speciesGentoo	578.629	75.362	7.678	0.000	430.391	726.867

Model Plots

library(ggplot2)
library(palmerpenguins)
library(dplyr)

penguins |>
  filter(!is.na(sex)) |>
  ggplot(aes(bill_length_mm, body_mass_g, colour = species)) +
  geom_point(alpha = 0.4) +
  geom_smooth(method = "lm", se = FALSE) +
  labs(
    x = "Bill length (mm)", y = "Body mass (g)", colour = "Species",
    title = "Penguin body mass vs bill length",
    subtitle = "Lines show fitted regression per species"
  ) +
  theme_minimal()

Model Plots

When to use tables vs. plots

• Use a table when you want to show exact values or the accuracy of the values is important to convey.
• Plots are often more useful for communicating a particular message, trend or pattern.
• You can also combine tables with visualisations!

Case study: heatmap table with formattable

library(dplyr)
library(tibble)
library(formattable)
mtcars |>
  select(mpg, disp, wt, hp) |>
  cor() |>
  as.data.frame() |>
  rownames_to_column("Variables") |>
  formattable(list(area(col = 2:5) ~ color_tile("#F5B7B1", "#7DCEA0")))

Case study: heatmap table with formattable

Variables	mpg	disp	wt	hp
mpg	1.0000000	-0.8475514	-0.8676594	-0.7761684
disp	-0.8475514	1.0000000	0.8879799	0.7909486
wt	-0.8676594	0.8879799	1.0000000	0.6587479
hp	-0.7761684	0.7909486	0.6587479	1.0000000

Making tables with {gt}

Packages for making tables in R

Table packages in R
Package	Output	Strength
knitr::kable	HTML / PDF / Word	Simple, zero dependencies
kableExtra	HTML / PDF	Extensions to kable
formattable	HTML	Inline colour and bar formatting
gt	HTML / PDF / Word	Layered API, highly customisable
gtsummary	HTML / PDF / Word	Automatic summary and model tables
DT	HTML	Interactive, sortable and filterable

Tip

We will focus on {gt} — it has a layered API that generalises to HTML, PDF, and Word, and is actively maintained by Posit.

Components of a table

The parts (roughly from top to bottom) are:

the Table Header (optional; with a title and possibly a subtitle) the Stub and the Stub Head (optional; contains row labels, optionally within row groups having row group labels and possibly summary labels when a summary is present) the Column Labels (contains column labels, optionally under spanner column labels) the Table Body (contains columns and rows of cells) the Table Footer (optional; possibly with footnotes and source notes)

The gt workflow

Start with a data frame, pipe into gt(), then add formatting and annotation layers one at a time:

gt(df) |>
  fmt_number(columns = where(is.numeric), decimals = 1) |>
  cols_label(col_name = "Human Label") |>
  cols_align(align = "right", columns = where(is.numeric)) |>
  tab_header(title = "Title", subtitle = "Subtitle") |>
  tab_source_note(md("Source: ...")) |>
  tab_footnote(footnote = "Note.", locations = cells_column_labels())

Example: Data sample

Show the first few rows of a dataset with clean labels:

library(dplyr)
library(gt)
library(palmerpenguins)
penguins |>
  slice_sample(n = 6) |>
  gt()

Example: Data sample

species	island	bill_length_mm	bill_depth_mm	flipper_length_mm	body_mass_g	sex	year
Gentoo	Biscoe	49.5	16.2	229	5800	male	2008
Adelie	Biscoe	36.5	16.6	181	2850	female	2008
Gentoo	Biscoe	49.1	14.8	220	5150	female	2008
Gentoo	Biscoe	51.1	16.5	225	5250	male	2009
Adelie	Dream	39.6	18.1	186	4450	male	2008
Gentoo	Biscoe	55.1	16.0	230	5850	male	2009

Example: Data sample

Improve the column names:

library(dplyr)
library(gt)
library(palmerpenguins)
penguins |>
  slice_sample(n = 6) |>
  gt() |>
  cols_label(
    bill_length_mm    = html("Bill length<br>(mm)"),
    bill_depth_mm     = html("Bill depth<br>(mm)"),
    flipper_length_mm = html("Flipper<br>(mm)"),
    body_mass_g       = html("Body mass<br>(g)")
  )

Example: Data sample

species	island	Bill length (mm)	Bill depth (mm)	Flipper (mm)	Body mass (g)	sex	year
Gentoo	Biscoe	48.7	14.1	210	4450	female	2007
Chinstrap	Dream	53.5	19.9	205	4500	male	2008
Chinstrap	Dream	50.6	19.4	193	3800	male	2007
Chinstrap	Dream	51.3	18.2	197	3750	male	2007
Adelie	Biscoe	41.1	19.1	188	4100	male	2008
Adelie	Torgersen	38.5	17.9	190	3325	female	2009

Example: Data sample

Now add formatting and supporting text

library(dplyr)
library(gt)
library(palmerpenguins)
penguins |>
  slice_sample(n = 6) |>
  gt() |>
  cols_label(
    bill_length_mm    = html("Bill length<br>(mm)"),
    bill_depth_mm     = html("Bill depth<br>(mm)"),
    flipper_length_mm = html("Flipper<br>(mm)"),
    body_mass_g       = html("Body mass<br>(g)")
  ) |>
  fmt_number(where(is.numeric), decimals = 1) |>
  tab_header(title = "Palmer Penguins — data sample")

Example: Data sample

Palmer Penguins — data sample
species	island	Bill length (mm)	Bill depth (mm)	Flipper (mm)	Body mass (g)	sex	year
Chinstrap	Dream	54.2	20.8	201.0	4,300.0	male	2,008.0
Gentoo	Biscoe	48.5	14.1	220.0	5,300.0	male	2,008.0
Chinstrap	Dream	55.8	19.8	207.0	4,000.0	male	2,009.0
Adelie	Dream	42.3	21.2	191.0	4,150.0	male	2,007.0
Adelie	Dream	33.1	16.1	178.0	2,900.0	female	2,008.0
Gentoo	Biscoe	50.5	15.2	216.0	5,000.0	female	2,009.0

Example: Summary statistics

Summarise first, then pipe into gt():

penguins |>
  group_by(species) |>
  summarise(
    n = n(),
    mean_bill = mean(bill_length_mm, na.rm = TRUE),
    sd_bill   = sd(bill_length_mm,   na.rm = TRUE)
  ) |>
  gt() |>
  fmt_number(c(mean_bill, sd_bill), decimals = 2) |>
  cols_label(
    species   = "Species",
    n         = md("*n*"),
    mean_bill = html("Mean bill<br>(mm)"),
    sd_bill   = html("SD bill<br>(mm)")
  ) |>
  tab_header(title = "Bill length by species")

Example: Summary statistics

Bill length by species
Species	n	Mean bill (mm)	SD bill (mm)
Adelie	152	38.79	2.66
Chinstrap	68	48.83	3.34
Gentoo	124	47.50	3.08

Example: Grouping Columns

Use tab_spanner() to group related columns under a shared header:

penguins |>
  mutate(year = as.factor(year)) |>
  tidyr::drop_na() |>
  group_by(species, sex) |>
  summarise(across(where(is.numeric), \(x) mean(x , na.rm = TRUE)),
            .groups = "drop") |>
  gt() |>
  cols_label(
    bill_length_mm    = html("Bill length<br>(mm)"),
    bill_depth_mm     = html("Bill depth<br>(mm)"),
    flipper_length_mm = html("Flipper<br>(mm)"),
    body_mass_g       = html("Body mass<br>(g)")
  ) |>
  tab_spanner(label = "Body Measurements", columns = c(bill_length_mm, bill_depth_mm, flipper_length_mm, body_mass_g)) |>
  fmt_number(where(is.numeric), decimals = 1) |>
  cols_align(align = "center", columns = -c(species, sex))

Example: Grouping Columns

species	sex	Body Measurements
		Bill length (mm)	Bill depth (mm)	Flipper (mm)	Body mass (g)
Adelie	female	37.3	17.6	187.8	3,368.8
Adelie	male	40.4	19.1	192.4	4,043.5
Chinstrap	female	46.6	17.6	191.7	3,527.2
Chinstrap	male	51.1	19.3	199.9	3,939.0
Gentoo	female	45.6	14.2	212.7	4,679.7
Gentoo	male	49.5	15.7	221.5	5,484.8

Example: Table head and footer

library(dplyr)
library(gt)
library(palmerpenguins)

penguins |>
  group_by(species) |>
  summarise(
    n         = n(),
    mean_bill = mean(bill_length_mm, na.rm = TRUE),
    sd_bill   = sd(bill_length_mm,   na.rm = TRUE),
    mean_mass = mean(body_mass_g,    na.rm = TRUE)
  ) |>
  gt() |>
  fmt_number(c(mean_bill, sd_bill, mean_mass), decimals = 1) |>
  cols_label(
    species   = "Species",
    n         = md("*n*"),
    mean_bill = html("Mean bill<br>(mm)"),
    sd_bill   = html("SD bill<br>(mm)"),
    mean_mass = html("Mean mass<br>(g)")
  ) |>
  cols_align(align = "right", columns = where(is.numeric)) |>
  tab_header(
    title    = "Penguin measurements by species",
    subtitle = "Chinstrap penguins have the longest bills on average"
  ) |>
  tab_source_note(md("Source: `palmerpenguins` package"))

Example: Table head and footer

Penguin measurements by species
Chinstrap penguins have the longest bills on average
Species	n	Mean bill (mm)	SD bill (mm)	Mean mass (g)
Adelie	152	38.8	2.7	3,700.7
Chinstrap	68	48.8	3.3	3,733.1
Gentoo	124	47.5	3.1	5,076.0
Source: palmerpenguins package

Table design

• Right-align numbers, left-align text, centre-align spanner labels
• Put units in column headers, not in cells (mm, g, %)
• Choose decimal places deliberately — match the precision that matters
• Group related columns with tab_spanner()

Texts accompanying tables

A table may be accompanied by:

• Title / header — names the table and its scope
• Caption — states the take-away message; why should the reader care?
• Footnotes — explain abbreviations, caveats, or exceptions in specific cells
• Source note — credits the data source
• Alt text — a plain-text description for screen readers (set via tbl-cap in Quarto)

Convention varies by audience and medium (academic paper, report, slide deck).

Adding table captions in Quarto

Markdown tables — caption is the last paragraph inside a ::: div with a tbl- label:

::: {#tbl-penguins}

| Species   | Bill (mm) |
| :-------- | --------: |
| Adelie    |      38.8 |
| Chinstrap |      48.8 |

Penguins measured at Palmer Station.
:::

Computational tables — use tbl-cap chunk option:

````{r}
#| label: tbl-penguins
#| tbl-cap: "Penguins measured at Palmer Station."

penguins |> gt()
````

Control caption position with tbl-cap-location: top (default for HTML) or bottom / margin.

The #tbl- label enables cross-referencing with @tbl-penguins. Caption location can also be set document-wide in YAML. See: https://quarto.org/docs/authoring/tables.html#caption-location

Communicating with Visualisation

Based on: https://cwd.numbat.space/week6/#/etc5523-title

Communication starts with intent

Before choosing chart type, or creating a custom visualisation, you need to know what you are trying to communicate. You should answer:

• What relationship or observation about the data are you trying to show?
• Who is the audience — how much context do they bring?
• What would “no effect” or “no pattern” look like — and how would you show that?

Example: Birth Countries of Australian Residents

This example is based on Australian Census data from 2016 and 2021.

The data have been filtered to look at the top 5 foreign countries of birth (which are the same across both census years): England, New Zealand, China, India and Philippines:

2016
Country	Count	%
England	907,570	3.9
New Zealand	518,466	2.2
China	509,555	2.2
India	455,389	1.9
Philippines	232,386	1.0

2021
Country	Count	%
England	927,490	3.6
India	673,352	2.6
China	549,618	2.2
New Zealand	530,492	2.1
Philippines	293,892	1.2

Top 5 overseas countries of birth, Australian census

Do you notice a change between 2016-2021?

Matching chart type to messages

Let’s try a different chart type to show our message more clearly.

Goal	Chart type
Compare values across categories	Bar chart (horizontal preferred)
Show distribution of one variable	Histogram, density, boxplot, violin
Show relationship between two variables	Scatter plot, line chart
Show composition or part-of-whole	Stacked bar, treemap
Show change over time	Line chart, area chart
Show many variables at once	Faceted plots, paired plots

Custom chart types

Common chart types might not highlight or convey the particular message or information you want. In this case, you might experiment with creating your own hybrid, composite or totally new chart types.

Message: India overtook China and New Zealand

Let’s try a line chart with two time points (2016 and 2021):

But notice, the legend and line order don’t match…

Can we improve clarity?

Let’s use direct labelling for each country line, and switch to percentages rather than absolute counts.

Adjusting labels?

Let’s use data() to only put labels in 2021, and adjust the y positions by modifying the percentage values so the labels don’t overlap.

birth_top5 |>
  arrange(census, birth) |>
  ggplot(aes(x = factor(census), y = percentage / 100, color = birth)) +
  geom_line(aes(group = birth)) +
  geom_point() +
  scale_y_continuous(labels = scales::percent) +
  geom_text(
    aes(label = birth),
    data = ~filter(.x, census == 2021) |>
      mutate(percentage = case_when(
        birth == "China"       ~ percentage + 0.05,
        birth == "New Zealand" ~ percentage - 0.05,
        TRUE                   ~ percentage
      )),
    hjust = 0, nudge_x = 0.05
  ) +
  labs(y = "Percentage of Australian residents", x = "Census year",
       caption = "Data source: Australian Census 2016 and 2021",
       title = "Top 5 countries of birth outside Australia") +
  theme(plot.title.position = "plot", legend.position = "none")

Adjusting labels?

Principles for Visualisation Design

What makes for a good visualisation?

• Are you showing the information needed to support your message? (e.g. statements about counts vs. percentages?)
• Is more important information highlighted visually? (e.g. bar -> line chart in birth place example)
• Are extraneous or distracting information removed or given lower visual priority? (e.g. filtering observations, removing duplicate labels)
• How easily could someone guess the title, subtitle, or caption of your visualisation?

Precision vs. Perception

Similar to the deciding between tables and plots, there are also trade-offs between designing for:

• quick insights and pattern recognition (perception) or
• precise measurements and exact readings (precision).

Source: Mastering Data Visualisation: Understanding the Hierarchy of Visual Cues (based on Cairo (2016))

Extension: Perceptual Tasks & Hierarchy

Perceptual tasks are elementary visual operations used to extract information from a chart:

• Position along a common scale
• Length comparison
• Angle estimation
• Area perception
• Colour saturation

Not all tasks are equally accurate — they form a ranked perceptual hierarchy:

• Cleveland and McGill (1984) — first empirical ranking
• Heer and Bostock (2010) — crowdsourced replication & extension
• Cairo (2016) — practical design synthesis

Position judgements are most precise; area and colour least.

Note

We will not cover specifics (that’s a whole separate course). A good starting article for practical advice: https://www.emrecanokten.com/articles/choosing-between-perception-or-precision-data-visualisation

Supporting text for visualisations

Support text helps readers interpret a visualisation without having to infer from data alone:

• Title & subtitle — what the chart shows; subtitle states the finding, not just the axes
• Axis labels — variable name and units; omit if clear from context
• Legend — maps visual encodings (colour, shape) to categories; prefer direct labels when possible
• Annotations & data labels — highlight specific values or observations inline
• Caption — directs viewer to the key takeaways and provides supplementary information (e.g. data source)
• Alt text — a text description for screen readers; convey the message, not just chart structure

Fixed marks with annotate()

annotate() places text, shapes, or arrows at fixed coordinates — not tied to any data row.

library(palmerpenguins)

penguins |>
  ggplot(aes(bill_length_mm, bill_depth_mm,
             colour = species)) +
  geom_point(alpha = 0.5, show.legend = FALSE) +
  annotate("rect",
    xmin = 40, xmax = 60,
    ymin = 12.5, ymax = 17.5,
    alpha = 0.15, fill = "steelblue") +
  annotate("text",
    x = 57, y = 13,
    label = "Gentoo", hjust = 0,
    fontface = "bold") +
  labs(x = "Bill length (mm)",
       y = "Bill depth (mm)")

Reference lines

geom_vline() / geom_hline() mark thresholds or events using a data-free layer.

museum_visitors |>
  ggplot(aes(x = jahr, y = wert, color = auspraegung)) +
  geom_line() +
  geom_vline(xintercept = 2020,
    linetype = "dashed", colour = "red") +
  annotate("text",
    x = 2020.3, y = 1000000,
    label = "COVID-19", hjust = 0,
    color = "red") +
  scale_y_continuous(
    labels = scales::label_comma()) +
  labs(x = "Year", y = "Annual visitors",
       title = "Visitors to Munich Museums",
       color = "Museum")

Captions

• Orient immediately — the opening sentence states the subject, variables, and context so readers understand the topic without looking at the axes
• Tell the story, not just the label — state the finding or trend (“Western region outpaced all others by 2023”), not just “Bar chart of sales by region”
• Include the key takeaway — surface the main conclusion; don’t make readers do the interpretive work themselves
• Provide essential context — time period, units, sample size, geographic scope, and important caveats (e.g. “excludes outliers above 99th percentile”)
• Explain non-obvious design choices — note log scales, non-standard encodings, or unusual chart types so readers decode correctly
• Credit the source — include a “Source:” line or footnote for data provenance
• Be concise but complete — two to four sentences; match vocabulary to the audience’s expertise

Alt-Text Principles

• Read the caption first — alt text should complement, not duplicate it: if the caption states the insight, focus alt text on structure; if the caption is generic, include the key insight; together they should give complete understanding
• Always state the chart type — e.g. “A scatter plot of…”; screen readers have no visual context
• Name variables, units, and ranges — axis variables, measurement units, and numerical bounds (e.g. “bill depth (mm), range 13–22”)
• Describe visual encodings — explain colour, size, or shape mappings (e.g. “colour indicates species”)
• Describe observable patterns per group — shape, range, or trend for each category, without adding interpretation beyond what’s visible
• Keep it to 1–3 sentences — be complete but concise; alt text is not a paragraph of prose

Example: Caption & Alt-Text

library(palmerpenguins)
penguins |>
  ggplot(aes(x = bill_depth_mm, y = bill_length_mm,
             color = species)) +
  geom_point() +
  scale_x_continuous(breaks = seq(13, 22, 2)) +
  scale_y_continuous(breaks = seq(30, 60, 10)) +
  labs(
    x = "Bill depth (mm)",
    y = "Bill length (mm)",
    colour = "Species"
  ) +
  theme(aspect.ratio = 1)

Example: Caption & Alt-Text (cont.)

Caption

Examining the differences in bill sizes of three species of penguins. Colour indicates species. Gentoo bills are smaller in depth but longer. Chinstrap and Adelie bills have similarly large depth but Chinstrap bills are longer.

Caption purpose

Orients sighted readers and states the key takeaway — assumes the reader can see the chart.

Alt text

A scatter plot of penguin bill measurements from the palmerpenguins package. Bill depth (mm) is on the x-axis (range 13–22 mm) and bill length (mm) is on the y-axis (range 30–60 mm). Three species are shown in different colours. Gentoo penguins cluster at low bill depth and high bill length. Adelie penguins cluster at high bill depth and lower bill length. Chinstrap penguins overlap with Adelie in bill length but have similarly large bill depth. All three groups show a positive within-group association between depth and length.

Alt text purpose

Full substitute for the image — conveys all information a sighted reader would perceive, including variable names, units, and per-group patterns.

Note

For more examples see: https://numbats.github.io/alt-text-for-data-plots/example.html

Tools for creating visualisations

Customising ggplot2

• look into ggplot2 documentation
• search for examples online and on GitHub (especially #tidytuesday)
• ask an LLM
• read R Graphics Cookbook by Winston Chang — practical recipes for common customisations
• follow ggplot2 release notes for new features

Choosing extension packages

Pick packages that are widely used and actively maintained (check GitHub activity and CRAN release dates). Prefer CRAN-published packages over GitHub-only ones — CRAN submission checks catch common issues and improve reproducibility.

ggplot2 ecosystem

Composite Plots

Modifying text in ggplot2

library(palmerpenguins)
penguins |>
  ggplot(aes(bill_length_mm, bill_depth_mm, colour = species)) +
  geom_point(alpha = 0.5) +
  theme_minimal() +
  labs(
    title    = "Palmer Station penguins",
    subtitle = "Chinstrap penguins have the longest bills on average",
    x        = "Bill length (mm)",
    y        = "Bill depth (mm)",
    caption  = "Source: Palmer Station LTER / palmerpenguins package",
    colour   = "Species"
  )

• The subtitle states the finding — not just what the axes show
• y = NULL removes a redundant axis label when the title makes it clear
• The caption credits the data source

Direct labelling

Replace a legend with labels placed directly on the data:

Direct labelling

Replace a legend with labels placed directly on the data:

library(ggrepel)
library(palmerpenguins)

penguins |>
  ggplot(aes(bill_length_mm, bill_depth_mm, colour = species)) +
  geom_point(show.legend = FALSE) +
  geom_text_repel(
    data = ~slice_max(.x, bill_length_mm, by = species, n = 1),
    aes(label = species),
    nudge_x = 2, show.legend = FALSE
  ) +
  labs(x = "Bill length (mm)", y = "Bill depth (mm)")

Reduces eye-travel between legend and data — especially useful when there are ≤ 5 groups.

Alt-Text Generation

Set alt text per figure with the fig-alt chunk option:

```{r}
#| fig-alt: "Line chart showing India's share rising from 4.5% to 6.1%..."
ggplot2(...) +
  geom_line(...) +
```

TIP: Drafting alt text with an LLM

Emil Hvitfeldt shows how to use Claude Code to auto-generate fig-alt for all figures in a Quarto document in one pass.

See: emilhvitfeldt.com/post/claude-code-alt-text-quarto

Remember to always review and adjust the alt text to make sure it matches your message and plot.

Summary

Communicating with Tables

• Choose between tables and plots based on your message and audience – which matter more? pattern or trend or exact values?
• Descriptive summaries can cover: variable list, observation count, missing values, distributions, and correlations
• Model result tables show estimates, uncertainty (SE or CI), and a fit statistics — use broom::tidy() to get a tidy data frame first
• Correlation tables and plots: {corrr} produces tidy correlation data frames ready to pipe into gt() or rplot()
• Cross-tabulations show joint frequency distributions — count() + pivot_wider() + gt() is the standard pipeline
• Combine tables and visualisations when you need both precision and pattern: formattable colour tiles are one lightweight option

Making tables with {gt}

• The gt workflow: wrangle first, then gt(), then stack formatting (fmt_*), column (cols_*), and annotation (tab_*) layers
• Column labels should carry units (mm, g, %) — keep cells clean with cols_label() and html() / md() helpers
• tab_spanner() groups related columns under a shared header
• Right-align numbers, left-align text; choose decimal places to match the precision that matters to your reader
• Every table should have a title and a source note;
• Captions should state the take-away, not just repeat the title
• Alt text for tables is set via tbl-cap in Quarto — write what a screen reader needs to understand the finding

Communicating with Visualisation

• Start with intent!
  • what relationship do you want to show?
  • who is the audience?
• Iterative improvement
  • start with a default chart, then adjust encodings, labels, and scale to foreground the message
  • does your chart match your message?
  • consider perceptual tasks and hierarchy as a guide
• Supporting text
  • The subtitle should state the finding, not just repeat the axes; the caption credits the data source
  • Alt text (fig-alt chunk option) conveys the chart’s message — not its structure — for screen readers

References

Cairo, Alberto. 2016. The Truthful Art: Data, Charts, and Maps for Communication. 1st ed. Indianapolis: New Riders.

Cleveland, William S., and Robert McGill. 1984. “Graphical Perception: Theory, Experimentation, and Application to the Development of Graphical Methods.” Journal of the American Statistical Association 79 (387): 531–54. https://doi.org/10.1080/01621459.1984.10478080.

Heer, Jeffrey, and Michael Bostock. 2010. “Crowdsourcing Graphical Perception: Using Mechanical Turk to Assess Visualization Design.” In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems, 203–12. Atlanta Georgia USA: ACM. https://doi.org/10.1145/1753326.1753357.