Open-Source Lunch Bites Webinar - Graphics Basics (Parts 1 & 2)

Lunch Bites
Author

PSI AIMS

Published

April 2, 2026

Overview

These notes cover the first two Graphics Basics sessions (6th March and 2nd April 2026), which introduced the foundations of data visualisation with ggplot2: what the package is and how it works, the layered grammar of graphics, core geometric objects, aesthetic mappings, grouping, colours, axes, and theme customisation. The sessions used a small simulated ADSL and ADLB dataset to build plots piece by piece.

Why visualise data?

Before writing a single line of code, it helps to ask three questions:

  • What is the purpose of the figure?
  • What is the context in which it will be seen?
  • Who is the audience?

Good figures present data clearly and honestly. The Information is Beautiful graphic continuum is a useful reference for thinking about which chart type suits a given data structure.

The ggplot2 package

ggplot2 is built on the Grammar of Graphics: you provide the data, describe how variables map to visual properties (aesthetics), and choose a geometric representation. The package handles the rest.

A typical plot is assembled in layers:

library(ggplot2)

# Core structure
ggplot(data = my_data, aes(x = var1, y = var2)) +
  geom_point()

The + operator combines layers. The order matters: later layers are drawn on top of earlier ones.

Two golden rules

  1. Start simple — get a basic plot working before adding anything.
  2. Build piece by piece — add one layer or argument at a time and check the result.

This approach makes it much easier to spot where something has gone wrong.

The figure framework

A ggplot2 figure is built from a small set of components:

  • Data — the data frame being plotted
  • aes() — aesthetic mappings (which variable goes on which axis, what controls colour, group, linetype, and so on)
  • Geom — the geometric object that represents the data (point, line, boxplot, bar, and so on)
  • Scales — control axis limits, breaks, labels, and colour palettes
  • Theme — controls the overall look: background, grid lines, fonts, legend position

Setting up data

The session used a small simulated subject-level dataset (dadsl) and a lab dataset (dadlb), created entirely in R without reading any external files.

set.seed(123)

n_subj <- 20

dadsl <- data.frame(
  SUBJID   = sprintf("SUBJ%03d", 1:n_subj),
  WEIGHTBL = round(rnorm(n_subj, mean = 75, sd = 12), 1),
  HEIGHTBL = round(rnorm(n_subj, mean = 170, sd = 10), 1),
  TRT01P   = sample(c("Placebo", "Drug A"), n_subj, replace = TRUE),
  TRTDURD  = sample(30:180, n_subj, replace = TRUE)
)

n_visits <- 5

dadlb <- do.call(
  rbind,
  lapply(dadsl$SUBJID, function(id) {
    data.frame(
      SUBJID  = id,
      PARAMCD = "ALT",
      ADY     = c(1, 7, 14, 28, 56),
      AVAL    = round(rlnorm(n_visits, meanlog = log(30), sdlog = 0.4), 1)
    )
  })
)

A first plot

Start by picking data and mapping variables to the x and y axes. Nothing is drawn until a geom is added.

# Canvas only – no geometric layer yet
ggplot(data = dadsl, aes(x = WEIGHTBL, y = HEIGHTBL))

# Add a point layer
ggplot(data = dadsl, aes(x = WEIGHTBL, y = HEIGHTBL)) +
  geom_point()

Geometric layers

Changing the geom changes how the data are represented. Layers can be combined.

# Line plot
ggplot(data = dadsl, aes(x = WEIGHTBL, y = HEIGHTBL)) +
  geom_line()

# Box plot
ggplot(data = dadsl, aes(x = TRT01P, y = TRTDURD)) +
  geom_boxplot()

# Combining layers – box plot with individual points on top
ggplot(data = dadsl, aes(x = TRT01P, y = TRTDURD)) +
  geom_boxplot() +
  geom_point()

# Adding a reference line
ggplot(data = dadsl, aes(x = WEIGHTBL, y = HEIGHTBL)) +
  geom_point() +
  geom_hline(yintercept = 200)

Moving data and aesthetics into the geom

Aesthetic mappings and data can be placed inside individual geoms rather than in the top-level ggplot() call. This is useful when different layers use different datasets.

# Data in the geom
ggplot() +
  geom_point(data = dadsl, aes(x = TRT01P, y = TRTDURD))

# Different data in each geom
ggplot() +
  geom_boxplot(data = dadsl, aes(x = TRT01P, y = TRTDURD)) +
  geom_point(
    data = dadsl %>% filter(SUBJID == "SUBJ007"),
    aes(x = TRT01P, y = 150)
  )

Making changes at the data level

It is often cleaner to prepare the data before passing it to ggplot2 than to fight the plot into shape after.

library(dplyr)

# Create a missing value where height is above 180
dadsl2 <- dadsl %>%
  mutate(WEIGHTBL = ifelse(HEIGHTBL > 180, NA, HEIGHTBL))

ggplot(data = dadsl2, aes(x = WEIGHTBL, y = HEIGHTBL)) +
  geom_point()

# Remove the missing row to suppress the warning message
dadsl3 <- dadsl2 %>%
  select(WEIGHTBL, HEIGHTBL) %>%
  filter(!is.na(WEIGHTBL))

ggplot(data = dadsl3, aes(x = WEIGHTBL, y = HEIGHTBL)) +
  geom_point()

Colours, shapes, linetypes, and transparency

Static appearance properties (not mapped to a variable) are set inside the geom as named arguments, outside of aes().

# Coloured squares
ggplot(data = dadsl2, aes(x = WEIGHTBL, y = HEIGHTBL)) +
  geom_point(colour = "red", shape = "square plus", size = 4)

# Custom line colours and types
ggplot(data = dadsl2, aes(x = WEIGHTBL, y = HEIGHTBL)) +
  geom_line(colour = "#CD6090", linetype = "longdash", linewidth = 1.5) +
  geom_hline(yintercept = 200, colour = "#96CDCD", linetype = "dotted", linewidth = 4)

# Transparent fill on a box plot
ggplot(data = dadsl, aes(x = TRT01P, y = TRTDURD)) +
  geom_boxplot(fill = "purple", alpha = 0.3)

When an appearance property depends on a variable in the data, it goes inside aes() instead.

Grouping

Grouping controls which observations are connected or drawn together.

# Fill by treatment group – automatically creates a legend
ggplot(data = dadsl, aes(x = TRT01P, y = TRTDURD, fill = TRT01P)) +
  geom_boxplot()

# Prepare a small lab dataset
dadlb2 <- dadlb %>%
  filter(SUBJID %in% c("SUBJ001", "SUBJ004", "SUBJ007"))

# Without grouping – all points joined up in visit order
ggplot(data = dadlb2, aes(x = ADY, y = AVAL)) +
  geom_line()

# Group by subject – separate lines, but no visual distinction
ggplot(data = dadlb2, aes(x = ADY, y = AVAL, group = SUBJID)) +
  geom_line()

# Linetype mapped to subject – distinct lines and a legend
ggplot(data = dadlb2, aes(x = ADY, y = AVAL, linetype = SUBJID)) +
  geom_line()

Grouping with manual colours

scale_colour_manual() lets you specify exact colours for each group and control the legend title.

ggplot(data = dadlb2, aes(x = ADY, y = AVAL, colour = SUBJID)) +
  geom_line(linewidth = 1.5) +
  scale_colour_manual(
    values = c("#0072B2", "#D55E00", "#CC79A7"),
    name = "Subject"
  )

Controlling axes

scale_x_continuous() and scale_y_continuous() control limits, tick positions, axis labels, and the small expansion gap at the edges.

ggplot(data = dadlb2, aes(x = ADY, y = AVAL, colour = SUBJID)) +
  geom_line(linewidth = 1.5) +
  scale_colour_manual(values = c("#0072B2", "#D55E00", "#CC79A7"), name = "Subject") +
  scale_x_continuous(
    limits = c(0, 60),
    breaks = seq(0, 60, 10),
    expand = c(0.01, 0.01),
    name = "Study Day"
  ) +
  scale_y_continuous(
    limits = c(10, 70),
    breaks = seq(10, 70, 10),
    expand = c(0.01, 0.01),
    name = "ALT Result"
  )

For discrete axes, scale_x_discrete() allows label replacement without changing the underlying data.

ggplot(data = dadsl, aes(x = TRT01P, y = TRTDURD)) +
  geom_boxplot() +
  scale_x_discrete(
    labels = c("Placebo" = "P", "Drug A" = "A"),
    name = "Treatment"
  )

Built-in themes

ggplot2 provides several complete themes that change the overall appearance of a plot.

theme_bw()
theme_linedraw()
theme_light()
theme_dark()
theme_minimal()
theme_classic()
theme_void()
theme_test()

Apply them by adding to the plot: + theme_bw().

Modifying themes

Individual theme elements can be overridden with theme(). Changes accumulate on top of whatever base theme is active.

ggplot(data = dadlb2, aes(x = ADY, y = AVAL, colour = SUBJID)) +
  geom_line(linewidth = 1.5) +
  scale_colour_manual(values = c("#0072B2", "#D55E00", "#CC79A7"), name = "Subject") +
  scale_x_continuous(limits = c(0, 60), breaks = seq(0, 60, 10),
                     expand = c(0.01, 0.01), name = "Study Day") +
  scale_y_continuous(limits = c(10, 70), breaks = seq(10, 70, 10),
                     expand = c(0.01, 0.01), name = "ALT Result") +
  theme(
    text = element_text(size = 10),
    legend.position = "bottom",
    legend.background = element_blank(),
    legend.box.background = element_rect(colour = "black"),
    panel.grid.minor.x = element_blank(),
    panel.grid.minor.y = element_blank(),
    plot.margin = unit(c(0.25, 0.75, 0.25, 0.75), "inches")
  )

Combining plots with patchwork

The patchwork package allows multiple ggplot2 figures to be assembled into a single layout.

library(patchwork)

f1 <- ggplot(data = dadlb2, aes(x = ADY, y = AVAL, colour = SUBJID)) +
  geom_line(linewidth = 1.5) +
  scale_colour_manual(values = c("#0072B2", "#D55E00", "#CC79A7"), name = "Subject") +
  scale_x_continuous(limits = c(0, 60), breaks = seq(0, 60, 10),
                     expand = c(0.01, 0.01), name = "Study Day") +
  scale_y_continuous(limits = c(10, 70), breaks = seq(10, 70, 10),
                     expand = c(0.01, 0.01), name = "ALT Result") +
  theme(
    text = element_text(size = 10),
    legend.position = "bottom",
    legend.background = element_blank(),
    legend.box.background = element_rect(colour = "black"),
    panel.grid.minor.x = element_blank(),
    panel.grid.minor.y = element_blank()
  )

f2 <- ggplot() +
  geom_boxplot(data = dadsl, aes(x = TRT01P, y = TRTDURD))

# Side by side
f1 + f2

# With relative widths
f1 + f2 + plot_layout(widths = c(1, 2))

Saving figures

ggsave() saves the most recent plot, or a named plot object, to a file. The ragg package provides higher-quality PNG output and is worth using for publication or regulatory submissions.

# Standard save
ggsave("f3.png", f3, device = png, width = 9.00, height = 4.75, dpi = 300)

# Higher quality with ragg
library(ragg)
ggsave("f3r.png", f3, device = agg_png, width = 9.00, height = 4.75, dpi = 300, scaling = 0.7)

The scaling argument in agg_png controls the relative size of text and elements — lower values make text larger relative to the plot area.

Graphics basics recap

Key points from the session:

  • Ask about purpose, context, and audience before starting a figure
  • ggplot2 builds plots from layers using +
  • Start simple and add one element at a time
  • Data and aesthetics can live in ggplot() or inside individual geoms
  • Static appearance properties go outside aes(); data-driven ones go inside it
  • scale_* functions control axes and colour palettes
  • theme() adjusts the non-data appearance of a plot
  • patchwork combines multiple ggplot2 figures
  • Use ragg for high-quality saved output

Resources

ggplot2

patchwork