Mastering Data Visualization with R

Great Court, The University of Queensland

Introduction

This tutorial introduces data visualisation with R, focusing on the ggplot2 package. It covers a wide range of plot types suited to different data structures and research questions — from scatter plots and distribution plots to Likert scale visualisations, heatmaps, time series, and publication-ready figures. Throughout, the emphasis is on choosing the right visualisation for a given question, understanding the grammar of graphics that underlies ggplot2, and developing the habits that lead to clear, reproducible, and honest data communication.

The tutorial works through a concrete dataset on preposition frequencies in historical English texts, providing a continuous research narrative that connects the individual examples. Exercises at the end of each section consolidate understanding.

Learning Objectives

By the end of this tutorial you will be able to:

1. Explain the grammar of graphics and how it structures ggplot2 code
2. Choose an appropriate visualisation type for a given data structure and research question
3. Create scatter plots, density plots, histograms, ridge plots, boxplots, violin plots, bar plots, heatmaps, line graphs, and ribbon plots in ggplot2
4. Visualise Likert scale survey data using grouped bar plots and gglikert
5. Customise plots with themes, colour palettes, labels, and annotations
6. Apply accessibility principles including redundant encoding and colourblind-safe palettes
7. Combine multiple plots into a single figure using patchwork
8. Save publication-quality figures in appropriate formats and resolutions
9. Avoid common visualisation mistakes including truncated axes, chartjunk, and overplotting

Prerequisite Tutorials

Before working through this tutorial, you should be familiar with:

• Getting Started with R
• Loading, Saving, and Generating Data in R
• Handling Tables in R

Citation

Martin Schweinberger. 2026. Mastering Data Visualization with R. The Language Technology and Data Analysis Laboratory (LADAL), The University of Queensland, Australia. url: https://ladal.edu.au/tutorials/data_viz_advanced/data_viz_advanced.html (Version 3.1.1).

---

Setup and Preparation

Section Overview

What you will learn: Which packages are needed and why; how to load the tutorial dataset; and how to set up a consistent colour palette for use throughout the tutorial

Installing required packages

Run this code once to install all required packages. It may take a few minutes.

install.packages("dplyr")
install.packages("stringr")
install.packages("ggplot2")
install.packages("tidyr")
install.packages("scales")
install.packages("ggridges")
install.packages("ggstats")
install.packages("ggstatsplot")
install.packages("EnvStats")
install.packages("likert")
install.packages("vcd")
install.packages("hexbin")
install.packages("patchwork")    # Combining multiple plots
install.packages("viridis")      # Colourblind-safe palettes
install.packages("flextable")
install.packages("devtools")

# Install ggflags from GitHub (country flags in plots)
devtools::install_github("jimjam-slam/ggflags")

Loading packages

library(dplyr)
library(stringr)
library(ggplot2)
library(tidyr)
library(flextable)
library(hexbin)
library(patchwork)
library(ggflags)
library(ggstats)
library(ggridges)
library(EnvStats)
library(scales)
library(viridis)

Loading and inspecting the data

We work throughout this tutorial with a dataset on preposition frequencies in historical English texts from the Penn Parsed Corpora of Historical English (PPCME, PPCEME, PPCMBE). Each row represents one text, and the key variables are described below.

pdat <- base::readRDS("tutorials/data_viz_advanced/data/pvd.rda", "rb")

Date	Genre	Text	Prepositions	Region	GenreRedux	DateRedux
1,736	Science	albin	166.01	North	NonFiction	1700-1799
1,711	Education	anon	139.86	North	NonFiction	1700-1799
1,808	PrivateLetter	austen	130.78	North	Conversational	1800-1913
1,878	Education	bain	151.29	North	NonFiction	1800-1913
1,743	Education	barclay	145.72	North	NonFiction	1700-1799
1,908	Education	benson	120.77	North	NonFiction	1800-1913
1,906	Diary	benson	119.17	North	Conversational	1800-1913
1,897	Philosophy	boethja	132.96	North	NonFiction	1800-1913
1,785	Philosophy	boethri	130.49	North	NonFiction	1700-1799
1,776	Diary	boswell	135.94	North	Conversational	1700-1799
1,905	Travel	bradley	154.20	North	NonFiction	1800-1913
1,711	Education	brightland	149.14	North	NonFiction	1700-1799
1,762	Sermon	burton	159.71	North	Religious	1700-1799
1,726	Sermon	butler	157.49	North	Religious	1700-1799
1,835	PrivateLetter	carlyle	124.16	North	Conversational	1800-1913

Variable descriptions:

• Date — year the text was written (continuous)
• Genre — text genre (Fiction, Legal, Religious, etc.)
• Text — source text identifier
• Prepositions — relative frequency of prepositions per 1,000 words
• Region — geographic origin of the text (North/South)
• GenreRedux — simplified genre categories (5 levels)
• DateRedux — time period categories (1150–1499, 1500–1599, etc.)

Setting up a colour palette

Using a consistent colour palette across all visualisations creates a coherent, professional look and reduces the cognitive load of switching between colour schemes. We define five colours here that we will reuse throughout.

clrs <- c("purple", "gray80", "lightblue", "orange", "gray30")

Colour resources

• R Color Reference — all named colours in R
• ColorBrewer — palettes designed for maps and data visualisation, many colourblind-safe
• Viridis — perceptually uniform, colourblind-safe palettes

For accessibility, prefer palettes from the viridis package or scale_color_brewer() with "Set2" or "Dark2".

---

Part 1: The Grammar of Graphics

Section Overview

What you will learn: The conceptual framework underlying ggplot2; the seven components of every plot; and how to read and write ggplot2 code systematically

Why ggplot2?

ggplot2 is the dominant data visualisation package in R for good reason. It is based on a coherent theoretical framework — the grammar of graphics — that makes it possible to construct any plot from a small set of building blocks. Rather than memorising individual plot functions, you learn a system: once you understand the grammar, you can build plots you have never seen before by composing components in new ways.

The grammar of graphics, formalised by Wilkinson (2005) and implemented in ggplot2 by Wickham (2010), describes a plot as the result of mapping data to aesthetics through geometric objects, with additional components controlling scales, coordinate systems, facets, and themes.

The seven components

Every ggplot2 plot is built from up to seven components:

1. Data — the data frame containing the variables to be visualised. Passed as the first argument to ggplot().

2. Aesthetics (aes()) — the mapping from data variables to visual properties: which variable goes on the x-axis, which on the y-axis, which controls colour, size, shape, transparency, and so on. Aesthetics defined inside ggplot() apply to all layers; aesthetics inside a specific geom_*() apply only to that layer.

3. Geometries (geom_*()) — the geometric objects used to represent the data. Points, lines, bars, boxes, ribbons, tiles, and text are all geometries. Each geom_*() call adds a new layer to the plot.

4. Scales (scale_*()) — control how aesthetic mappings are translated into visual properties. For example, scale_color_manual() specifies exact colours; scale_x_log10() log-transforms the x-axis; scale_y_continuous(labels = scales::percent) formats y-axis labels as percentages.

5. Facets (facet_wrap(), facet_grid()) — split the data into subplots by the values of one or more categorical variables. Faceting is one of the most powerful features of ggplot2 for comparing patterns across groups.

6. Coordinate system (coord_*()) — controls the space in which the plot is drawn. coord_flip() swaps x and y; coord_polar() creates polar (circular) coordinates; coord_cartesian() sets axis limits without dropping data points.

7. Theme (theme_*(), theme()) — controls all non-data visual elements: background colour, gridlines, font sizes, axis tick marks, legend position, and so on. theme_bw() and theme_minimal() are good defaults for publication work.

The ggplot2 template

Every ggplot2 call follows this template:

ggplot(data = <DATA>, aes(x = <X>, y = <Y>, color = <GROUP>)) +
  geom_<TYPE>(<PARAMETERS>) +
  scale_<AESTHETIC>_<TYPE>(<PARAMETERS>) +
  facet_<TYPE>(vars(<VARIABLE>)) +
  coord_<TYPE>() +
  theme_<STYLE>() +
  labs(title = "<TITLE>", x = "<X LABEL>", y = "<Y LABEL>")

The + operator adds layers and components to the plot. The order generally does not matter for the final result, but it is conventional to put data layers first, then scales, then facets, then theme, then labels.

Reading existing ggplot2 code

When you encounter unfamiliar ggplot2 code, read it layer by layer. Ask: what data is being used? What is mapped to x, y, colour, and other aesthetics? What geometric objects are being drawn? What scales and themes have been applied? This decomposition makes even complex plots understandable.

It controls which variables are mapped to which axes.
It specifies the type of geometric object used to represent the data.
It controls all non-data visual elements of the plot, such as background colour, gridlines, font sizes, axis labels, and legend position.
It determines how data values are transformed before plotting.

---

Part 2: Exploring Relationships

Section Overview

What you will learn: Scatter plots as the foundation for showing relationships between two continuous variables; adding colour, shape, and trend lines; using facets; managing overplotting with transparency, density contours, and hex plots

Scatter plots

Scatter plots are the most direct way to visualise the relationship between two continuous variables. Each point represents one observation.

When to use: Two continuous variables; sample size small enough that individual points can be seen (roughly < 5,000 without overplotting strategies).

Basic scatter plot

ggplot(data = pdat,
       aes(x = Date,
           y = Prepositions)) +
  geom_point() +
  theme_bw() +
  labs(x = "Year",
       y = "Prepositions per 1,000 words")

Reading the code

• ggplot() initialises the plot and sets the default data and aesthetics
• aes(x = Date, y = Prepositions) maps the variable Date to the x-axis and Prepositions to the y-axis
• geom_point() adds a layer of points — one per row in the data
• theme_bw() applies a clean black-and-white theme
• labs() sets axis labels

Adding colour and shape

Using both colour and shape to encode the same variable is called redundant encoding. It makes plots more accessible: readers who cannot distinguish colours (about 8% of men have some form of colour vision deficiency) can still use the shapes, and the plot retains its meaning when printed in greyscale.

ggplot(pdat,
       aes(Date, Prepositions,
           color = GenreRedux,
           shape = GenreRedux)) +
  geom_point(size = 2) +
  scale_shape_manual(name = "Genre", values = 1:5) +
  scale_color_manual(name = "Genre", values = clrs) +
  theme_bw() +
  theme(legend.position = "top") +
  labs(x = "Year", y = "Prepositions per 1,000 words")

Faceted scatter plots with trend lines

When points from multiple groups overlap, faceting into separate panels makes individual group patterns visible. Adding a trend line with geom_smooth() makes the overall direction of change within each group explicit.

ggplot(pdat, aes(Date, Prepositions, color = Genre)) +
  facet_wrap(vars(Genre), ncol = 4) +
  geom_point(alpha = 0.4) +
  geom_smooth(method = "lm", se = FALSE, linewidth = 0.8) +
  theme_bw() +
  theme(
    legend.position = "none",
    axis.text.x = element_text(size = 8, angle = 90)
  ) +
  labs(x = "Year", y = "Prepositions per 1,000 words")

Facets: when to use them

Facets work best when you have 3–8 groups whose within-group patterns are the focus, and when direct across-group value comparison is less important than seeing each group’s trend clearly. Avoid facets when groups need to be directly overlaid for comparison, or when you have more than about 10 groups.

Managing overplotting

When many points occupy the same region, individual points become invisible. Three strategies address this:

Transparency (alpha) — making points semi-transparent so density is visible as colour intensity.

2D density contours (geom_density_2d) — contour lines showing where data is concentrated, like a topographic map.

Hex plots (geom_hex) — the plotting region is divided into hexagonal bins; each bin is coloured by the number of points it contains. Effective for very large datasets.

ggplot(pdat, aes(x = Date, y = Prepositions, color = GenreRedux)) +
  facet_wrap(vars(GenreRedux), ncol = 5) +
  geom_density_2d() +
  theme_bw() +
  theme(
    legend.position = "none",
    axis.text.x = element_text(size = 8, angle = 90)
  ) +
  labs(x = "Year", y = "Prepositions per 1,000 words")

pdat |>
  ggplot(aes(x = Date, y = Prepositions)) +
  geom_hex() +
  scale_fill_gradient(low = "lightblue", high = "darkblue",
                      name = "Count") +
  theme_bw() +
  labs(x = "Year", y = "Prepositions per 1,000 words",
       title = "Hex plot: point density")

Approach	Best for	Limitation
Points	Small–medium datasets, seeing all data	Gets cluttered with many points
Transparency	Moderate overplotting	Still unclear at very high density
Density contours	Showing concentration patterns	Harder to interpret than points
Hex bins	Very large datasets	Requires comparable x–y scales

---

Part 3: Showing Distributions

Section Overview

What you will learn: Density plots, histograms, ridge plots, boxplots, and violin plots — when each is appropriate and what each reveals that the others do not

Density plots

Density plots show the estimated probability density of a continuous variable as a smooth curve. They are particularly useful for comparing the shape of a distribution across groups.

ggplot(pdat, aes(Date, fill = Region)) +
  geom_density(alpha = 0.5) +
  scale_fill_manual(values = clrs[1:2]) +
  theme_bw() +
  theme(legend.position = c(0.1, 0.9)) +
  labs(x = "Year", y = "Density",
       title = "Temporal distribution of texts by region")

The plot shows that southern texts continue into the 1800s while northern texts end around 1700, with a period of overlap in between.

Histograms

Histograms divide a continuous variable into equal-width bins and count how many observations fall in each. Unlike density plots, they show actual counts and make the discretisation of the data explicit.

ggplot(pdat, aes(Prepositions)) +
  geom_histogram(bins = 30, fill = "steelblue", color = "white") +
  theme_bw() +
  labs(title = "Distribution of preposition frequencies",
       x = "Prepositions per 1,000 words",
       y = "Count")

Histogram vs. bar plot

A histogram shows the distribution of one continuous variable. The bins are ranges of values, and there are no gaps between bars (the variable is continuous).

A bar plot shows counts or values for discrete categories. Bars are separated by gaps to reflect the categorical (not continuous) nature of the x-axis.

Confusing the two is one of the most common plotting mistakes in student work.

Ridge plots

Ridge plots (also called joy plots) show offset density curves for multiple groups, making it easy to compare shapes across many groups simultaneously. They are particularly effective when you have more groups than can comfortably be shown in overlapping densities.

pdat |>
  ggplot(aes(x = Prepositions, y = GenreRedux, fill = GenreRedux)) +
  geom_density_ridges() +
  theme_ridges() +
  theme(legend.position = "none") +
  labs(y = "", x = "Relative frequency of prepositions per 1,000 words",
       title = "Preposition frequency distributions by genre")

Boxplots

Boxplots display five summary statistics simultaneously: the median (line inside the box), the first and third quartiles (the box edges, enclosing the interquartile range, IQR), and the whiskers extending to 1.5 times the IQR beyond each box edge. Points beyond the whiskers are plotted individually as potential outliers.

ggplot(pdat, aes(DateRedux, Prepositions, fill = DateRedux)) +
  geom_boxplot() +
  scale_fill_manual(values = clrs) +
  theme_bw() +
  theme(legend.position = "none") +
  labs(x = "Time period", y = "Prepositions per 1,000 words")

Notched boxplots

Adding notch = TRUE draws notches around the median. If notches of two boxes do not overlap, there is strong visual evidence that the medians differ significantly. This is a useful quick check, though it is not a substitute for formal statistical testing.

ggplot(pdat, aes(DateRedux, Prepositions, fill = DateRedux)) +
  geom_boxplot(notch = TRUE,
               outlier.colour = "red",
               outlier.shape = 2,
               outlier.size = 3) +
  scale_fill_manual(values = clrs) +
  theme_bw() +
  theme(legend.position = "none") +
  labs(x = "Time period", y = "Prepositions per 1,000 words",
       title = "Notched boxplots: overlapping notches suggest similar medians")

Enhanced boxplots with jittered points

Overlaying the individual data points on the boxplot reveals the sample size and distribution simultaneously.

ggplot(pdat, aes(DateRedux, Prepositions, fill = DateRedux, color = DateRedux)) +
  geom_boxplot(varwidth = TRUE, color = "black", alpha = 0.3) +
  geom_jitter(alpha = 0.3, height = 0, width = 0.2) +
  facet_grid(~Region) +
  EnvStats::stat_n_text(y.pos = 65) +
  theme_bw() +
  theme(legend.position = "none") +
  labs(x = "", y = "Frequency per 1,000 words",
       title = "Preposition use across time and regions",
       subtitle = "Box width proportional to sample size; n shown below each box")

Violin plots

Violin plots mirror a density plot on both sides of a central axis, giving them their characteristic shape. They show the full distribution shape — including multimodality — while remaining compact enough to compare across groups.

ggplot(pdat, aes(DateRedux, Prepositions, fill = DateRedux)) +
  geom_violin(trim = FALSE, alpha = 0.5) +
  scale_fill_manual(values = clrs) +
  theme_bw() +
  theme(legend.position = "none") +
  labs(x = "Time period", y = "Prepositions per 1,000 words",
       title = "Violin plots reveal distribution shape")

Choosing between distribution plot types

Plot type	Reveals	Best for	Avoid when
Histogram	Counts in bins	Single variable, showing counts	Comparing many groups
Density	Smooth shape	Comparisons, overlapping groups	Exact counts needed
Ridge	Multiple shapes	Many groups (> 4)	Fewer than 3 groups
Boxplot	Five-number summary + outliers	Statistical summaries	Distribution shape matters
Violin	Shape + summary	Detecting multimodality	Very small samples

A histogram, because it shows exact counts and is the most familiar plot type.
A boxplot, because it always shows outliers clearly.
A violin plot, because it shows both the distribution shape (like a density plot) and summary statistics, and can reveal multimodal distributions that a boxplot would hide.
A ridge plot, because it handles multiple groups better than any other option.

---

Part 4: Categorical Data

Section Overview

What you will learn: Bar plots in their basic, grouped, stacked, and normalised forms; Likert scale visualisation; and the case against pie charts

Bar plots

Bar plots show counts, frequencies, or summary values for categorical groups. They are the workhorse of categorical data visualisation.

First, we create summary data:

bdat <- pdat |>
  dplyr::mutate(DateRedux = factor(DateRedux)) |>
  group_by(DateRedux) |>
  dplyr::summarise(Frequency = n()) |>
  dplyr::mutate(Percent = round(Frequency / sum(Frequency) * 100, 1))

bdat

# A tibble: 5 × 3
  DateRedux Frequency Percent
  <fct>         <int>   <dbl>
1 1150-1499        34     6.3
2 1500-1599       180    33.5
3 1600-1699       225    41.9
4 1700-1799        53     9.9
5 1800-1913        45     8.4

Basic bar plot

ggplot(bdat, aes(DateRedux, Percent, fill = DateRedux)) +
  geom_bar(stat = "identity") +
  geom_text(aes(y = Percent - 3,
                label = paste0(Percent, "%")),
            color = "white", size = 4) +
  scale_fill_manual(values = clrs) +
  theme_bw() +
  theme(legend.position = "none") +
  labs(x = "Time period",
       y = "Percentage of documents",
       title = "Distribution of texts across time periods")

stat = "identity" explained

geom_bar() defaults to stat = "count", which counts the number of rows per group. When your data already contains the values to plot — as bdat$Percent does here — use stat = "identity" to plot the values as given without any additional aggregation.

Grouped and stacked bar plots

ggplot(pdat, aes(Region, fill = DateRedux)) +
  geom_bar(position = position_dodge(), stat = "count") +
  scale_fill_manual(values = clrs) +
  theme_bw() +
  labs(x = "Region", y = "Number of documents", fill = "Time period",
       title = "Document counts by region and time period (grouped)")

ggplot(pdat, aes(DateRedux, fill = GenreRedux)) +
  geom_bar(stat = "count") +
  scale_fill_manual(values = clrs) +
  theme_bw() +
  labs(x = "Time period", y = "Number of documents", fill = "Genre",
       title = "Genre composition across time periods (stacked)")

ggplot(pdat, aes(DateRedux, fill = GenreRedux)) +
  geom_bar(stat = "count", position = "fill") +
  scale_fill_manual(values = clrs) +
  scale_y_continuous(labels = scales::percent) +
  theme_bw() +
  labs(x = "Time period", y = "Proportion of documents", fill = "Genre",
       title = "Relative genre composition over time (100% stacked)")

Bar type	Use when
Basic / grouped	Comparing absolute counts across groups
Stacked	Showing composition and total simultaneously
100% normalised	Only proportions matter, not absolute counts

Likert scale visualisations

Survey data recorded on Likert scales (e.g. Strongly Disagree to Strongly Agree) requires careful visualisation because the response categories are ordered, the neutral midpoint is meaningful, and the visual emphasis should reflect valence.

ldat <- base::readRDS("tutorials/data_viz_advanced/data/lid.rda", "rb")
head(ldat)

   Course Satisfaction
1 Chinese            1
2 Chinese            1
3 Chinese            1
4 Chinese            1
5 Chinese            1
6 Chinese            1

Grouped bar plot

nlik <- ldat |>
  dplyr::group_by(Course, Satisfaction) |>
  dplyr::summarize(Frequency = n(), .groups = "drop")

ggplot(nlik, aes(Satisfaction, Frequency, fill = Course)) +
  geom_bar(stat = "identity", position = position_dodge()) +
  scale_fill_manual(values = clrs[1:3]) +
  geom_text(aes(label = Frequency),
            vjust = 1.6, color = "white",
            position = position_dodge(0.9), size = 3.5) +
  scale_x_discrete(
    limits = 1:5,
    labels = c("Very\nDissatisfied", "Dissatisfied",
               "Neutral", "Satisfied", "Very\nSatisfied")
  ) +
  theme_bw() +
  labs(title = "Student satisfaction by course",
       x = "Satisfaction level", y = "Number of students")

Cumulative distribution plot

ggplot(ldat, aes(x = Satisfaction, color = Course)) +
  geom_step(aes(y = after_stat(y)), stat = "ecdf", linewidth = 1.5) +
  scale_colour_manual(values = clrs[1:3]) +
  scale_x_discrete(
    limits = 1:5,
    labels = c("Very\nDissatisfied", "Dissatisfied",
               "Neutral", "Satisfied", "Very\nSatisfied")
  ) +
  theme_bw() +
  labs(title = "Cumulative satisfaction distribution",
       y = "Cumulative proportion", x = "Satisfaction level")

Reading cumulative distribution plots

A steeper slope at any point means responses are concentrated in that range. A line that runs high on the left means many dissatisfied respondents. When two lines cross, it means the distributions have different shapes — one group may have more extreme responses in both directions.

gglikert: diverging bar chart

The gglikert() function from the ggstats package creates diverging stacked bar charts that place negative responses on the left and positive responses on the right, with neutral in the middle. This is currently considered the most effective visualisation for Likert data.

sdat <- base::readRDS("tutorials/data_viz_advanced/data/sdd.rda", "rb")

colnames(sdat)[3:ncol(sdat)] <- paste0(
  "Q", str_pad(1:10, 2, "left", "0"), ": ",
  colnames(sdat)[3:ncol(sdat)]
) |>
  stringr::str_replace_all("\\.", " ") |>
  stringr::str_squish() |>
  stringr::str_replace_all("$", "?")

lbs <- c("Disagree", "Somewhat\nDisagree", "Neutral",
         "Somewhat\nAgree", "Agree")

survey <- sdat |>
  dplyr::mutate_if(is.character, factor) |>
  dplyr::mutate_if(is.numeric, factor, levels = 1:5, labels = lbs) |>
  drop_na() |>
  as.data.frame()

survey |>
  dplyr::select(matches("01|02|03|04")) |>
  gglikert(labels_size = 2.5, add_labels = FALSE) +
  ggtitle("Survey responses: selected questions") +
  scale_fill_brewer(palette = "RdBu")

Likert visualisation best practices

• Keep response categories in their natural order — never sort by frequency
• Use a diverging colour palette (e.g. red–blue) centred on the neutral midpoint
• Show the neutral category separately in the middle of the bar
• Include sample sizes when comparing groups
• Prefer diverging bar charts over plain stacked bars for communication

Pie charts: use with caution

The case against pie charts

Human visual perception is much better at comparing lengths (bar plot) than angles or areas (pie chart). Research consistently shows that people make more accurate judgements from bar charts than from pie charts, especially when slices are of similar size or when there are more than three categories.

Pie charts may be acceptable when there are only two or three categories and one clearly dominates. In most other situations, a bar chart communicates more accurately.

piedata <- bdat |>
  dplyr::arrange(desc(DateRedux)) |>
  dplyr::mutate(Position = cumsum(Percent) - 0.5 * Percent)

p_bar <- ggplot(bdat, aes("", Percent, fill = DateRedux)) +
  geom_bar(stat = "identity", position = position_dodge(), width = 0.7) +
  scale_fill_manual(values = clrs) +
  theme_minimal() +
  labs(title = "Bar plot", y = "Percent", x = "")

p_pie <- ggplot(piedata, aes("", Percent, fill = DateRedux)) +
  geom_bar(stat = "identity", width = 1, color = "white") +
  coord_polar("y", start = 0) +
  scale_fill_manual(values = clrs) +
  theme_void() +
  geom_text(aes(y = Position, label = paste0(Percent, "%")),
            color = "white", size = 4) +
  labs(title = "Pie chart")

p_bar + p_pie

Without looking at the percentage labels, try to identify the second-largest category in each plot. The bar plot makes this easy; the pie chart makes it difficult.

A grouped bar plot, because it is the most common plot type for categorical data.
A pie chart for each time period, because pie charts are best for showing parts of a whole.
A 100% normalised stacked bar plot, because it directly shows how the proportions of each genre changed across periods while maintaining the correct total of 100% for each period.
A scatter plot, because it can show change over time on the x-axis.

---

Part 5: Advanced Visualisations

Section Overview

What you will learn: Heatmaps and association plots for matrix data; word clouds for text data; flag plots for international comparisons; dot plots with error bars; and diverging bar plots

Heatmaps

Heatmaps use colour intensity to represent values in a two-dimensional matrix. They are effective for showing patterns across many combinations of two categorical variables.

heatdata <- pdat |>
  dplyr::group_by(DateRedux, GenreRedux) |>
  dplyr::summarise(Prepositions = mean(Prepositions), .groups = "drop") |>
  tidyr::spread(DateRedux, Prepositions)

heatmx <- as.matrix(heatdata[, 2:5])
rownames(heatmx) <- heatdata$GenreRedux
heatmx_scaled <- scale(heatmx)

heatmap(heatmx_scaled,
        scale  = "none",
        col    = colorRampPalette(c("blue", "white", "red"))(50),
        margins = c(7, 10),
        main   = "Preposition frequency: standardised mean by genre and period")

The dendrograms show which genres (rows) and time periods (columns) cluster together based on their preposition frequency profiles. Blue indicates below-average frequency; red indicates above-average frequency.

Association and mosaic plots

Association plots and mosaic plots from the vcd package visualise the relationship between two categorical variables, showing deviations from statistical independence.

library(vcd)

assocdata <- pdat |>
  dplyr::mutate(
    GenreRedux = dplyr::case_when(
      GenreRedux == "Conversational" ~ "Conv.",
      GenreRedux == "Religious"      ~ "Relig.",
      TRUE ~ GenreRedux
    )
  ) |>
  dplyr::group_by(GenreRedux, DateRedux) |>
  dplyr::summarise(Prepositions = round(mean(Prepositions), 0),
                   .groups = "drop") |>
  tidyr::spread(DateRedux, Prepositions)

assocmx <- as.matrix(assocdata[, 2:6])
rownames(assocmx) <- assocdata$GenreRedux

assoc(assocmx, shade = TRUE,
      main = "Association plot: genre by time period")

mosaic(assocmx, shade = TRUE, legend = TRUE,
       main = "Mosaic plot: genre composition over time")

Interpreting these plots:

• Bars or tiles above the baseline: more than expected under independence
• Bars or tiles below the baseline: less than expected
• Blue shading: significantly more than expected (p < 0.05)
• Red shading: significantly less than expected (p < 0.05)
• Bar width in the association plot: contribution to the chi-square statistic

Word clouds

Word clouds represent term frequencies visually, with word size proportional to frequency. They are visually engaging but imprecise — word sizes are difficult to compare accurately. Use them for exploratory purposes or presentations, not as primary evidence in a paper.

library(quanteda)
library(quanteda.textplots)

clinton <- base::readRDS("tutorials/data_viz_advanced/data/Clinton.rda", "rb") |>
  paste0(collapse = " ")
trump   <- base::readRDS("tutorials/data_viz_advanced/data/Trump.rda", "rb") |>
  paste0(collapse = " ")

corp_dom <- quanteda::corpus(c(clinton, trump))
attr(corp_dom, "docvars")$Author <- c("Clinton", "Trump")

dfm_dom <- corp_dom |>
  quanteda::tokens(remove_punct = TRUE) |>
  quanteda::tokens_remove(stopwords("english")) |>
  quanteda::dfm() |>
  quanteda::dfm_group(groups = corp_dom$Author) |>
  quanteda::dfm_trim(min_termfreq = 200, verbose = FALSE)

dfm_dom |>
  quanteda.textplots::textplot_wordcloud(
    comparison = TRUE,
    max_words  = 50,
    color      = c("blue", "red")
  )

Country flags in visualisations

The ggflags package allows country flags to be used as data point markers, making international comparisons more immediately readable.

flagsdf <- data.frame(
  Region  = c("Australia", "Canada", "Great Britain", "India",
               "Ireland", "New Zealand", "United States"),
  Percent = c(0.022, 0.017, 0.025, 0.010, 0.019, 0.020, 0.036),
  Kachru  = c("Inner circle", "Inner circle", "Inner circle", "Outer circle",
               "Inner circle", "Inner circle", "Inner circle"),
  country = c("au", "ca", "gb", "in", "ie", "nz", "us")
)

flagsdf |>
  ggplot(aes(x = reorder(Region, Percent),
             y = Percent,
             country = country,
             fill = Kachru)) +
  geom_bar(stat = "identity") +
  ggflags::geom_flag(size = 5) +
  geom_text(aes(label = scales::percent(Percent, accuracy = 0.1)),
            hjust = -0.3, size = 3) +
  coord_flip(ylim = c(0, 0.045)) +
  scale_fill_manual(values = c("lightblue", "coral")) +
  scale_y_continuous(labels = scales::percent) +
  theme_minimal() +
  labs(x = "", y = "Vulgar language percentage",
       title = "Vulgar language use by English-speaking region",
       fill = "English type") +
  theme(legend.position = c(0.8, 0.3),
        panel.grid.major = element_blank())

Dot plots with error bars

Dot plots showing means with confidence intervals are often preferable to bar plots for continuous outcomes because they avoid the visual distortion caused by showing the mean as the height of a bar that starts at zero.

ggplot(pdat, aes(x = reorder(Genre, Prepositions, mean),
                 y = Prepositions,
                 group = Genre)) +
  stat_summary(fun = mean, geom = "point", size = 4,
               aes(color = Genre)) +
  stat_summary(fun.data = mean_cl_boot, geom = "errorbar",
               width = 0.2, linewidth = 1) +
  coord_cartesian(ylim = c(80, 200)) +
  theme_bw(base_size = 12) +
  theme(axis.text.x = element_text(angle = 45, hjust = 1),
        legend.position = "none") +
  labs(x = "", y = "Prepositions per 1,000 words",
       title = "Mean preposition frequency by genre",
       subtitle = "Error bars show 95% bootstrap confidence intervals")

Diverging bar plots

Diverging bar plots show deviation from a reference value, with positive deviations extending in one direction and negative in the other. They are useful for comparing group profiles against a baseline.

Test1 <- c(11.2, 13.5, 200, 185, 1.3, 3.5)
Test2 <- c(12.2, 14.7, 210, 175, 1.9, 3.0)
Test3 <- c(13.2, 15.1, 177, 173, 2.4, 2.9)

testdata <- data.frame(Test1, Test2, Test3)
rownames(testdata) <- c(
  "Feature1_Student", "Feature1_Reference",
  "Feature2_Student", "Feature2_Reference",
  "Feature3_Student", "Feature3_Reference"
)

plottable <- data.frame(
  Test    = rep(rownames(t(testdata[1,] - testdata[2,])), 3),
  Value   = c(t(testdata[1,] - testdata[2,]),
              t(testdata[3,] - testdata[4,]),
              t(testdata[5,] - testdata[6,])),
  Feature = rep(c("Feature A", "Feature B", "Feature C"), each = 3)
)

ggplot(plottable, aes(Test, Value, fill = Test)) +
  facet_grid(vars(Feature), scales = "free_y") +
  geom_bar(stat = "identity") +
  geom_hline(yintercept = 0, linetype = "dashed", color = "red") +
  scale_fill_manual(values = clrs[1:3]) +
  theme_bw() +
  theme(legend.position = "none") +
  labs(x = "Test", y = "Deviation from reference",
       title = "Learner performance relative to native speaker reference",
       subtitle = "Positive = above reference; negative = below reference")

---

Part 6: Time Series and Line Graphs

Section Overview

What you will learn: Line graphs for discrete and continuous time variables; smoothed trend lines; ribbon plots for displaying uncertainty; and how to choose between these approaches

Basic line graphs

Line graphs connect data points in temporal order, making trends and trajectories visible. The group aesthetic tells ggplot2 which points to connect.

pdat |>
  dplyr::group_by(DateRedux, GenreRedux) |>
  dplyr::summarise(Frequency = mean(Prepositions), .groups = "drop") |>
  ggplot(aes(x = DateRedux, y = Frequency,
             group = GenreRedux,
             color = GenreRedux)) +
  geom_line(linewidth = 1.2) +
  geom_point(size = 3) +
  scale_color_manual(values = clrs) +
  theme_minimal() +
  labs(title = "Preposition frequency over time by genre",
       x = "Time period",
       y = "Mean frequency per 1,000 words",
       color = "Genre")

Smoothed line graphs

For continuous time variables with many data points, LOESS smoothing (locally estimated scatterplot smoothing) reveals the underlying trend while absorbing noise from individual observations.

ggplot(pdat, aes(x = Date, y = Prepositions,
                 color = GenreRedux,
                 linetype = GenreRedux)) +
  geom_smooth(se = FALSE, linewidth = 1.2) +
  scale_linetype_manual(
    values = c("solid", "dashed", "dotted", "dotdash", "longdash"),
    name = "Genre"
  ) +
  scale_colour_manual(values = clrs, name = "Genre") +
  theme_bw() +
  theme(legend.position = "top") +
  labs(x = "Year", y = "Relative frequency\nper 1,000 words",
       title = "Smoothed trends in preposition use (LOESS)")

Using both colour and line type (redundant encoding) keeps the lines distinguishable in greyscale and for readers with colour vision deficiency.

Ribbon plots: showing uncertainty

Ribbon plots (geom_ribbon) display ranges or intervals as shaded bands around a central line. They are effective for communicating uncertainty, variability, or the full range of observed values.

pdat |>
  dplyr::mutate(DateRedux = as.numeric(DateRedux)) |>
  dplyr::group_by(DateRedux) |>
  dplyr::summarise(
    Mean = mean(Prepositions),
    Min  = min(Prepositions),
    Max  = max(Prepositions),
    SD   = sd(Prepositions),
    .groups = "drop"
  ) |>
  ggplot(aes(x = DateRedux, y = Mean)) +
  geom_ribbon(aes(ymin = Min, ymax = Max),
              fill = "gray80", alpha = 0.3) +
  geom_ribbon(aes(ymin = Mean - SD, ymax = Mean + SD),
              fill = "lightblue", alpha = 0.4) +
  geom_line(linewidth = 1.2, color = "darkblue") +
  scale_x_continuous(labels = names(table(pdat$DateRedux))) +
  theme_minimal() +
  labs(title = "Preposition frequency: mean with variability",
       subtitle = "Dark blue = mean; light blue = ±1 SD; grey = full range",
       x = "Time period",
       y = "Frequency per 1,000 words")

geom_smooth() and geom_line() are interchangeable and produce identical results.
geom_smooth() uses statistical smoothing (LOESS or linear regression) to draw a trend curve, which reduces noise but does not show the actual data points. geom_line() connects the actual data points in order, showing every measured value but potentially hiding the overall trend in noisy data.
geom_smooth() is only for scatter plots; geom_line() is only for time series.
geom_line() shows uncertainty intervals automatically, while geom_smooth() does not.

---

Part 7: Combining Plots with patchwork

Section Overview

What you will learn: How to combine multiple ggplot2 plots into a single figure using the patchwork package; layout operators; adding shared titles, subtitles, and labels; and when combining plots is appropriate

Why combine plots?

A multi-panel figure is often more effective than a series of separate plots when:

• You want readers to compare related results side by side
• A single visualisation cannot show all the relevant aspects of the data
• You are preparing a figure for a publication that expects one figure file per result

The patchwork package provides a simple and powerful syntax for combining ggplot2 plots.

Basic patchwork syntax

The three main operators are:

• | — place plots side by side (horizontal)
• / — place plots one above the other (vertical)
• + — add to the current layout (follows row-by-row order)
• () — group plots for nested layouts

# Create three component plots
p1 <- ggplot(pdat, aes(x = DateRedux, y = Prepositions, fill = DateRedux)) +
  geom_boxplot() +
  scale_fill_manual(values = clrs) +
  theme_bw() +
  theme(legend.position = "none") +
  labs(x = "Time period", y = "Prepositions per 1,000 words",
       title = "A: Boxplots")

p2 <- ggplot(pdat, aes(x = Prepositions, y = GenreRedux, fill = GenreRedux)) +
  geom_density_ridges() +
  theme_ridges() +
  theme(legend.position = "none") +
  labs(x = "Prepositions per 1,000 words", y = "",
       title = "B: Ridge plot")

p3 <- pdat |>
  dplyr::group_by(DateRedux, GenreRedux) |>
  dplyr::summarise(Mean = mean(Prepositions), .groups = "drop") |>
  ggplot(aes(x = DateRedux, y = Mean,
             group = GenreRedux, color = GenreRedux)) +
  geom_line(linewidth = 1.1) +
  geom_point(size = 2.5) +
  scale_color_manual(values = clrs) +
  theme_minimal() +
  labs(x = "Time period", y = "Mean frequency",
       color = "Genre", title = "C: Line graph")

# Combine: p1 and p2 side by side, with p3 below
(p1 | p2) / p3

Shared labels and annotations

patchwork provides plot_annotation() for adding overall titles, subtitles, and captions, and plot_layout() for controlling spacing and shared legends.

(p1 | p2) / p3 +
  plot_annotation(
    title    = "Preposition frequency in historical English texts",
    subtitle = "Three complementary views of the same dataset",
    caption  = "Source: Penn Parsed Corpora of Historical English",
    tag_levels = "A"
  )

Collecting legends

When multiple plots share the same colour mapping, you can collect the legends into a single shared legend with plot_layout(guides = "collect").

pa <- ggplot(pdat, aes(DateRedux, Prepositions, fill = GenreRedux)) +
  geom_boxplot() +
  scale_fill_manual(values = clrs) +
  theme_bw() +
  labs(x = "Time period", y = "Prepositions", fill = "Genre")

pb <- ggplot(pdat, aes(DateRedux, fill = GenreRedux)) +
  geom_bar(position = "fill") +
  scale_fill_manual(values = clrs) +
  scale_y_continuous(labels = scales::percent) +
  theme_bw() +
  labs(x = "Time period", y = "Proportion", fill = "Genre")

pa2 <- pa + theme(legend.position = "bottom")
pb2 <- pb + theme(legend.position = "bottom")

pa2 | pb2

Use p1 + p2 + p3, which always arranges three plots in a single row.
patchwork cannot create layouts where one plot spans a full row below two side-by-side plots.
Use (p1 | p2) / p3, which places p1 and p2 side by side in the top row and p3 spanning the full width in the bottom row.
Use p1 / (p2 | p3), which places p3 below and p1 and p2 above — the same result.

---

Part 8: Publication-Ready Plots and Choosing Wisely

Section Overview

What you will learn: What makes a plot publication-ready; saving figures in the right format and resolution; colour accessibility; a decision framework for choosing plot types; and the most common visualisation mistakes to avoid

The anatomy of a publication-ready plot

A plot ready for a journal article or conference proceedings should have:

• A clear, informative title and (where appropriate) a subtitle
• Axis labels that name the variable and include units
• A legend that is necessary and clearly positioned
• A theme appropriate to the publication context (usually theme_bw() or theme_minimal() rather than the default grey background)
• Font sizes large enough to be legible at the final printed size
• A colourblind-accessible colour palette
• A caption noting the data source and what error bars or ribbons represent

Complete example

pdat |>
  dplyr::group_by(DateRedux, GenreRedux) |>
  dplyr::summarise(
    Mean = mean(Prepositions),
    SE   = sd(Prepositions) / sqrt(n()),
    N    = n(),
    .groups = "drop"
  ) |>
  ggplot(aes(x = DateRedux, y = Mean,
             color = GenreRedux, group = GenreRedux)) +
  geom_line(linewidth = 1.2) +
  geom_point(size = 3) +
  geom_errorbar(aes(ymin = Mean - SE, ymax = Mean + SE),
                width = 0.2, linewidth = 0.8) +
  scale_color_manual(
    name   = "Text genre",
    values = clrs,
    labels = c("Conversational", "Fiction", "Legal", "Non-fiction", "Religious")
  ) +
  scale_y_continuous(breaks = seq(100, 200, 20), limits = c(100, 200)) +
  theme_bw(base_size = 14) +
  theme(
    legend.position       = c(0.15, 0.65),
    legend.background     = element_rect(fill = "white", color = "black"),
    panel.grid.minor      = element_blank(),
    plot.title            = element_text(face = "bold", size = 16),
    plot.subtitle         = element_text(size = 12, color = "gray30"),
    plot.caption          = element_text(size = 10, hjust = 0)
  ) +
  labs(
    title    = "Historical trends in preposition usage",
    subtitle = "Analysis of English texts from 1150 to 1913",
    x        = "Time period",
    y        = "Mean frequency (per 1,000 words)",
    caption  = "Source: Penn Parsed Corpora of Historical English (PPC)\nError bars show ±1 SE"
  )

Saving figures

# For journal submission (300 dpi minimum)
ggsave("preposition_trends.png",  width = 10, height = 6, dpi = 300)

# For vector graphics (no resolution limit — scales to any size)
ggsave("preposition_trends.pdf",  width = 10, height = 6)

# For web use
ggsave("preposition_trends_web.png", width = 10, height = 6, dpi = 150)

Format guide

PNG — raster format; use for web, slides, and figures containing photographs. Specify dpi = 300 for print.

PDF — vector format; use for journal submission where possible. Scales to any size without loss of quality. Best for plots containing text and sharp geometric elements.

TIFF — some journals require TIFF. Use dpi = 600 for posters.

SVG — vector format; useful for web and for figures you may need to edit further in Inkscape or Illustrator.

Colour accessibility

Approximately 8% of men and 0.5% of women have some form of colour vision deficiency. Designing accessible plots benefits all readers, not only those with colour vision differences.

p_problem <- pdat |>
  dplyr::group_by(DateRedux, GenreRedux) |>
  dplyr::summarise(Mean = mean(Prepositions), .groups = "drop") |>
  ggplot(aes(DateRedux, Mean, fill = GenreRedux)) +
  geom_bar(stat = "identity", position = "dodge") +
  scale_fill_manual(values = c("red", "green", "blue", "yellow", "purple")) +
  ggtitle("Problematic colours") +
  theme_minimal() +
  theme(axis.text.x = element_text(angle = 90, vjust = 0.5, hjust = 1),
        legend.position = "none")

p_better <- pdat |>
  dplyr::group_by(DateRedux, GenreRedux) |>
  dplyr::summarise(Mean = mean(Prepositions), .groups = "drop") |>
  ggplot(aes(DateRedux, Mean, fill = GenreRedux)) +
  geom_bar(stat = "identity", position = "dodge") +
  scale_fill_viridis_d() +
  ggtitle("Colourblind-friendly (viridis)") +
  theme_minimal() +
  theme(axis.text.x = element_text(angle = 90, vjust = 0.5, hjust = 1),
        legend.position = "none")

p_problem | p_better

Colourblind-safe options in ggplot2:

• scale_color_viridis_d() / scale_fill_viridis_d() — for discrete variables
• scale_color_viridis_c() / scale_fill_viridis_c() — for continuous variables
• scale_color_brewer(palette = "Set2") or "Dark2" — ColorBrewer palettes, many colourblind-safe
• Redundant encoding (colour + shape, or colour + line type) as a complement

Choosing the right plot: a decision framework

By data structure

One continuous variable — show distribution:

• Small samples (< 50): dot plot, strip plot
• Medium samples (50–500): histogram, density plot
• Large samples (500+): density plot, violin plot
• Summary statistics: boxplot

One continuous + one categorical — compare groups:

• Distributions: boxplot, violin plot, ridge plot
• Means with uncertainty: dot plot with error bars
• Show all data: jittered points

Two continuous variables — show relationship:

• Basic: scatter plot
• Overplotting: hex plot, 2D density
• With trend: add geom_smooth()
• Groups: colour, shape, or facets

Two categorical variables — show association:

• Frequencies: grouped or stacked bar plot
• Proportions: 100% normalised bar, mosaic plot
• Statistical deviations: association plot

Time series — show change:

• Discrete time points: line graph with points
• Continuous time: smoothed line, ribbon plot
• Multiple series: coloured lines or small multiples

Three or more variables — multivariate:

• Third variable categorical: colour + facets
• Third variable continuous: colour gradient or bubble size
• Many variables: heatmap

Common mistakes to avoid

3D charts — almost never appropriate. They distort values through perspective effects and make precise comparison impossible. Use 2D plots with grouping, colour, or facets instead.

Dual y-axes — can be used to misrepresent relationships between variables by independently scaling each axis. Prefer faceted plots or normalising both variables to the same scale.

Truncated y-axis on bar plots — bar heights encode values by length from zero. Cutting the axis at a non-zero value exaggerates differences. Bar plots must start at zero. Dot plots with error bars can use a truncated axis because they do not encode values by length from a baseline.

Too many colours — more than about six colours becomes difficult to distinguish. Consider reducing categories, using facets, or highlighting one group while greying the rest.

Chartjunk — decorative elements (unnecessary gridlines, 3D shadows, background images, clipart) distract from the data and add no information. Start with theme_minimal() or theme_bw() and add only what is needed.

Sorting bars randomly — unless the categories have a natural order (time periods, scale levels), sort bars by value to make rank comparisons easy.

Yes, because the differences are real and the truncated axis makes them easier to see.
Yes, as long as the axis break is clearly labelled.
No. Bar plots encode values as heights measured from zero. Cutting the y-axis at 150 makes a difference of 20 units (160 vs 180) appear as a much larger proportion of the bar than it would if the axis started at zero. This visually exaggerates the difference and could mislead readers. The y-axis on a bar plot must start at zero. A dot plot with error bars could legitimately use a truncated axis because it does not encode values by distance from a baseline.
It depends on the journal's guidelines.

---

Final Challenge: Capstone Project

Comprehensive data visualisation project

You have learned all the core techniques. The capstone is to create a coherent data story using the pdat dataset (or your own data).

Required components:

1. At least three different plot types from different sections — one showing distributions, one showing relationships, and one showing categorical comparisons
2. Publication-ready quality: proper titles, labels and captions; a colourblind-friendly palette; appropriate themes; clear legends
3. At least one combined figure using patchwork with a shared annotation
4. A written narrative: a short introduction explaining your research question; brief transition text between plots explaining what each shows; and a conclusion summarising what the visualisations reveal

Example research questions to explore:

• How has genre composition changed across the historical periods covered in the corpus?
• Are there regional differences in preposition frequency, and do they interact with time period?
• Which genres show the greatest variability in preposition use, and what might this reflect about genre norms?

Suggested deliverables: A fully ggplot2::annotated .qmd document with all code, at least three saved publication-quality figures (PNG, 300 dpi), and a brief 2–3 sentence caption for each figure as it would appear in a paper.

---

Citation & Session Info

Citation

Martin Schweinberger. 2026. Mastering Data Visualization with R. The Language Technology and Data Analysis Laboratory (LADAL), The University of Queensland, Australia. url: https://ladal.edu.au/tutorials/data_viz_advanced/data_viz_advanced.html (Version 3.1.1), doi: 10.5281/zenodo.19332872.

@manual{martinschweinberger2026mastering,
  author       = {Martin Schweinberger},
  title        = {Mastering Data Visualization with R},
  year         = {2026},
  note         = {https://ladal.edu.au/tutorials/data_viz_advanced/data_viz_advanced.html},
  organization = {The Language Technology and Data Analysis Laboratory (LADAL), The University of Queensland, Australia},
  edition      = {3.1.1}
  doi      = {10.5281/zenodo.19332872}
}

sessionInfo()

R version 4.4.2 (2024-10-31 ucrt)
Platform: x86_64-w64-mingw32/x64
Running under: Windows 11 x64 (build 26200)

Matrix products: default


locale:
[1] LC_COLLATE=English_United States.utf8 
[2] LC_CTYPE=English_United States.utf8   
[3] LC_MONETARY=English_United States.utf8
[4] LC_NUMERIC=C                          
[5] LC_TIME=English_United States.utf8    

time zone: Australia/Brisbane
tzcode source: internal

attached base packages:
[1] grid      stats     graphics  grDevices datasets  utils     methods  
[8] base     

other attached packages:
 [1] quanteda.textplots_0.95 quanteda_4.2.0          vcd_1.4-13             
 [4] viridis_0.6.5           viridisLite_0.4.2       scales_1.4.0           
 [7] EnvStats_3.0.0          ggridges_0.5.6          ggstats_0.10.0         
[10] ggflags_0.0.4           patchwork_1.3.0         hexbin_1.28.5          
[13] flextable_0.9.11        tidyr_1.3.2             ggplot2_4.0.2          
[16] stringr_1.6.0           dplyr_1.2.0             checkdown_0.0.13       

loaded via a namespace (and not attached):
 [1] tidyselect_1.2.1        farver_2.1.2            S7_0.2.1               
 [4] fastmap_1.2.0           fontquiver_0.2.1        rpart_4.1.23           
 [7] XML_3.99-0.18           labelled_2.14.1         digest_0.6.39          
[10] lifecycle_1.0.5         cluster_2.1.6           magrittr_2.0.4         
[13] compiler_4.4.2          Hmisc_5.2-2             rlang_1.1.7            
[16] tools_4.4.2             utf8_1.2.6              yaml_2.3.10            
[19] data.table_1.17.0       grImport2_0.3-3         knitr_1.51             
[22] stopwords_2.3           askpass_1.2.1           labeling_0.4.3         
[25] htmlwidgets_1.6.4       xml2_1.3.6              RColorBrewer_1.1-3     
[28] foreign_0.8-87          withr_3.0.2             purrr_1.2.1            
[31] nnet_7.3-19             gdtools_0.5.0           colorspace_2.1-1       
[34] MASS_7.3-61             isoband_0.2.7           cli_3.6.5              
[37] rmarkdown_2.30          ragg_1.5.1              generics_0.1.4         
[40] rstudioapi_0.17.1       commonmark_2.0.0        splines_4.4.2          
[43] BiocManager_1.30.27     base64enc_0.1-6         vctrs_0.7.2            
[46] Matrix_1.7-2            jsonlite_2.0.0          fontBitstreamVera_0.1.1
[49] litedown_0.9            ISOcodes_2024.02.12     hms_1.1.4              
[52] htmlTable_2.4.3         Formula_1.2-5           jpeg_0.1-11            
[55] systemfonts_1.3.1       glue_1.8.0              codetools_0.2-20       
[58] stringi_1.8.7           gtable_0.3.6            lmtest_0.9-40          
[61] tibble_3.3.1            pillar_1.11.1           htmltools_0.5.9        
[64] openssl_2.3.2           R6_2.6.1                textshaping_1.0.0      
[67] evaluate_1.0.5          lattice_0.22-6          markdown_2.0           
[70] haven_2.5.4             backports_1.5.0         png_0.1-8              
[73] renv_1.1.7              fontLiberation_0.1.0    fastmatch_1.1-8        
[76] Rcpp_1.1.1              zip_2.3.2               uuid_1.2-1             
[79] checkmate_2.3.2         gridExtra_2.3           nlme_3.1-166           
[82] mgcv_1.9-1              officer_0.7.3           xfun_0.56              
[85] zoo_1.8-13              forcats_1.0.0           pkgconfig_2.0.3        

AI Transparency Statement

This tutorial was re-developed with the assistance of Claude (claude.ai), a large language model created by Anthropic. Claude was used to help revise the tutorial text, structure the instructional content, generate the R code examples, and write the checkdown quiz questions and feedback strings. All content was reviewed, edited, and approved by the author (Martin Schweinberger), who takes full responsibility for the accuracy and pedagogical appropriateness of the material. The use of AI assistance is disclosed here in the interest of transparency and in accordance with emerging best practices for AI-assisted academic content creation.

Back to top

Back to LADAL home

Resources and Further Reading

Books

• Wickham, H. (2016). ggplot2: Elegant Graphics for Data Analysis (2nd ed.). Springer. Free online: ggplot2-book.org
• Healy, K. (2018). Data Visualization: A Practical Introduction. Princeton University Press. Free online: socviz.co
• Wilke, C. O. (2019). Fundamentals of Data Visualization. O’Reilly. Free online: clauswilke.com/dataviz

Online tools and references

• R Graph Gallery — hundreds of examples with reproducible code
• Data to Viz — decision tree for choosing plot types
• ggplot2 documentation — full function reference
• ColorBrewer — palette design tool
• patchwork documentation — combining plots

Practice datasets

Built into R: mpg, diamonds, economics, midwest

From packages: palmerpenguins (palmerpenguins), gapminder (gapminder), nycflights13 (nycflights13)

---

Quick Reference

Common geoms

Geom	Use for
geom_point()	Scatter plots, dot plots
geom_line()	Line graphs, time series
geom_bar()	Bar plots (counts or values)
geom_boxplot()	Distribution summaries with outliers
geom_violin()	Distribution shapes
geom_histogram()	Single variable distribution (counts)
geom_density()	Smooth distribution curves
geom_smooth()	Trend lines and regression curves
geom_errorbar()	Confidence intervals, error bars
geom_ribbon()	Ranges, uncertainty bands
geom_tile()	Heatmaps (ggplot2 version)
geom_hex()	Hex bins for large scatter data
geom_density_2d()	2D concentration contours

Common aesthetics

Aesthetic	Controls
x, y	Axis position
color / colour	Border or line colour
fill	Interior fill colour
size	Point size or text size
linewidth	Line thickness (replaces size for lines)
shape	Point shape
alpha	Transparency (0 = invisible, 1 = opaque)
linetype	Line style (solid, dashed, dotted, etc.)
group	Which observations to connect (lines)

Common themes

Theme	Character
theme_bw()	White background, black borders — good for publication
theme_minimal()	Minimal; no background panel
theme_classic()	Classic axis lines, no gridlines
theme_void()	No axes or gridlines — for maps, etc.
theme_ridges()	Optimised for ridge plots

Position adjustments

Position	Use for
position_dodge()	Side-by-side bars
position_stack()	Stacked bars
position_fill()	100% normalised stacked bars
position_jitter()	Spread overlapping points
position_identity()	Plot values exactly as given