Chapter 4 Intro to ggplot

We’ll use the ICAN-ICAR 2025 survey data with variables such as:

  • ReadingIRTScore: Child’s reading latent ability score
  • MathIRTScore: Child’s maths latent ability score
  • ch02: Child’s age
  • ch03: Child’s gender
  • EnrolmentStatus: child’s school enrollment status
  • ch04a: whether the child has eye difficulties

Load the packages and load and prepare the data:

You don’t need to understand all the details of ican-icar-2025-v1 yet.
For now, just remember:

  • We’ll be using a data frame called dat.
  • Each row is a child.

4.1 What is ggplot2?

ggplot2 is the R package we’ll use to make graphs.

  • It’s part of the tidyverse (a family of R packages for data analysis).
  • It’s based on the Grammar of Graphics idea.
  • Instead of giving you a few pre-made plots, it lets you build plots from simple pieces (layers).
  • You can use it effectively without fully understanding the underlying theory (we’ll learn by doing).

In practice, we’ll keep this simple mental model:

To make a plot in ggplot2 we always say:
“Use this data, map these variables to the axes/colour/etc.,
and draw them with this geom (points, bars, lines, …).”

4.1.1 Optional: Grammar of Graphics (background)

If you’re curious about the theory:

  • Graphics = distinct layers (data, aesthetic mappings, geoms, …).
  • Aesthetic mapping: connect variables in the data (numbers, labels) to what we see (position, colour, size, shape).

Don’t worry if this feels abstract right now; the examples will make it concrete.

4.2 The main pieces of ggplot2

A ggplot2 plot is built from several elements.

For our intro, we only need these three:

  1. data – the data frame we’re plotting (here: dat).
  2. aes(…) – aesthetic mappings: how variables map to what we see (x-axis, y-axis, colour, size, shape).
  3. geom_…() – the geometric object that draws the data (points, lines, bars, etc.).

Later (optional), we can also use:

  • stats – automatic summaries (e.g. means, counts, smooth lines).
  • scales – axis ranges and colour scales.
  • coordinate systems – how axes are drawn.
  • facets – small multiples of the same plot.
  • themes – fonts, grid lines, background (“non-data ink”).

4.3 Building your first plot: data → aes → geom

We’ll start by plotting household expenditure against household income.

4.3.1 Step 1: Choose the data

First we tell ggplot2 which data frame to use:

ggplot(data = dat)

This creates a blank plotting area (a coordinate system), but we haven’t told it what to draw yet, so you’ll likely see an empty plot.

4.3.2 Step 2: Map variables with aes()

Next, we say which variables go to which axis using aes():

ggplot(data = dat,
       mapping = aes(x = ReadingIRTScore, y = MathIRTScore))

Read this as:

  • Put ReadingIRTScore on the x-axis (horizontal).
  • Put ch02 on the y-axis (vertical).

This still doesn’t draw points — it only defines the mapping.

4.3.3 Step 3: Add a geometry with geom_point()

Now we add a geometry to actually draw the data. For a scatterplot, we use geom_point():

ggplot(data = dat,
       mapping = aes(x = ReadingIRTScore, y = MathIRTScore)) +
  geom_point()
## Warning: Removed 1019 rows containing missing values or values outside the scale range (`geom_point()`).

Think of this as:

“Using data dat,
put reading score on x, maths score on y,
and draw one point per child.”

This data + aes + geom pattern is the core template you’ll reuse for almost every plot in ggplot2. You can find other geometries that might interest you, i.e., a bar graph is geom_bar(), a line graph is geom_line(), etc. Here, we keep it simple by honing a scatter plot to show many design controls you could attain from ggplot.

4.4 Aesthetics: colour, shape, size

aes() can also control things like colour, shape, and size.

Big idea:

Whatever you put inside aes(...) is controlled by the data.
For example, aes(colour = ch03) means “use the variable ch03 to decide the colour of each point”.

We’ll reuse our scatterplot and experiment.

4.4.1 Mapping shape to a variable

ggplot(data = dat,
       mapping = aes(x = ReadingIRTScore, 
                     y = MathIRTScore, 
                     shape = ch03)) +
  geom_point()
## Warning: Removed 1019 rows containing missing values or values outside the scale range (`geom_point()`).

  • Different gender groups (ch03) get different shapes.

4.4.2 Mapping colour to a variable

ggplot(data = dat,
       mapping = aes(x = ReadingIRTScore, 
                     y = MathIRTScore, 
                     color = ch03)) +
  geom_point()
## Warning: Removed 1019 rows containing missing values or values outside the scale range (`geom_point()`).

  • Different groups get different colours.
  • A legend is added automatically.

4.4.3 Mapping size to a variable (e.g., age)

ggplot(data = dat,
       mapping = aes(x = ReadingIRTScore, 
                     y = MathIRTScore, 
                     size = ch02)) +
  geom_point(alpha = 0.1)
## Warning: Removed 1019 rows containing missing values or values outside the scale range (`geom_point()`).

  • Children with higher ch02 get larger points.
  • alpha = 0.1 makes points more transparent, so dense regions are easier to see.

4.4.4 Using a categorical variable

If you have categories like EnrolmentStatus, you can map those to colour or fill:

dat = dat |>
       filter(!is.na(ReadingIRTScore), !is.na(MathIRTScore))
ggplot(data = dat,
       mapping = aes(x = ReadingIRTScore,
                     y = MathIRTScore,
                     colour = EnrolmentStatus)) +
  geom_point(alpha = 0.25)

  • Each BMI category gets a different colour.
  • Missing categories are filtered out with filter(!is.na(bmi.bins)).

4.5 Aesthetic mapping vs fixed settings (some nuance)

Sometimes you want a fixed colour for all points (not data-driven).

  • Inside aes() → the value comes from the data.
  • Outside aes() → you set a fixed value.

For example:

# Colour determined by data (race)
ggplot(data = dat,
       mapping = aes(x = ReadingIRTScore,
                     y = MathIRTScore,
                     colour = EnrolmentStatus)) +
  geom_point(alpha = 0.25)

# Fixed colour (all points blue)
ggplot(data = dat,
       mapping = aes(x = ReadingIRTScore,
                     y = MathIRTScore,
                     colour = EnrolmentStatus)) +
  geom_point(alpha = 0.25, colour = "steelblue")

This distinction becomes important when you combine multiple layers or want to control colours manually, but for now just note the pattern.

4.6 Small polish: labels and a simple theme

Let’s tidy up our core scatterplot a little bit:

library(scales)

ggplot(data = dat,
       mapping = aes(x = ReadingIRTScore,
                     y = MathIRTScore,
                     colour = EnrolmentStatus)) +
  geom_point(alpha = 0.25) +
  labs(
    x = "Reading latent ability score",
    y = "Maths latent ability score",
    colour = "School enrollment status",
    title = "Foundational Skills in Senegal",
    subtitle = "ICAN-ICAR 2025 Survey"
  ) +
  scale_x_continuous(labels = label_comma()) +
  scale_y_continuous(labels = label_comma()) +
  theme_classic()

Here we:

  • Use labs() to set axis labels, legend title, and plot title/subtitle.
  • Use theme_minimal() to clean up the background and gridlines.
  • Just tip: by default, ggplot sometimes shows big numbers in scientific notation. We can tell ggplot to use “normal looking” numbers with commas by adding scale_x_continuous() etc.

For a intro, this level of theme usage is enough.

4.7 Practice: Test your knowledge with mpg

The ggplot2 package comes with a built-in dataset called mpg.

Load it:

library(ggplot2)  # already loaded via tidyverse, but explicit is fine
library(skimr)

data(mpg)

# Optional: look at its structure
skim(mpg)
Table 4.1: Data summary
Name mpg
Number of rows 234
Number of columns 11
_______________________
Column type frequency:
character 6
numeric 5
________________________
Group variables None

Variable type: character

skim_variable n_missing complete_rate min max empty n_unique whitespace
manufacturer 0 1 4 10 0 15 0
model 0 1 2 22 0 38 0
trans 0 1 8 10 0 10 0
drv 0 1 1 1 0 3 0
fl 0 1 1 1 0 5 0
class 0 1 3 10 0 7 0

Variable type: numeric

skim_variable n_missing complete_rate mean sd p0 p25 p50 p75 p100 hist
displ 0 1 3.470000000000000195399 1.290000000000000035527 1.600000000000000088818 2.399999999999999911182 3.299999999999999822364 4.599999999999999644729 7 ▇▆▆▃▁
year 0 1 2003.500000000000000000000 4.509999999999999786837 1999.000000000000000000000 1999.000000000000000000000 2003.500000000000000000000 2008.000000000000000000000 2008 ▇▁▁▁▇
cyl 0 1 5.889999999999999680256 1.610000000000000097700 4.000000000000000000000 4.000000000000000000000 6.000000000000000000000 8.000000000000000000000 8 ▇▁▇▁▇
cty 0 1 16.859999999999999431566 4.259999999999999786837 9.000000000000000000000 14.000000000000000000000 17.000000000000000000000 19.000000000000000000000 35 ▆▇▃▁▁
hwy 0 1 23.440000000000001278977 5.950000000000000177636 12.000000000000000000000 18.000000000000000000000 24.000000000000000000000 27.000000000000000000000 44 ▅▅▇▁▁

You can use ?mpg or help(mpg) to see more information about the variables.

Using the template:

ggplot(data = <DATA>,
       mapping = aes(x = <X>, y = <Y>, ...)) +
  <GEOM_FUNCTION>()

try these:

  1. Use a scatterplot to show the relationship between displ (engine displacement, in litres) and hwy (highway miles per gallon)
  2. In the same plot, set the colour of the points to class
  3. Graph a boxplot of hwy by class.
  4. Graph a bar plot of class.

4.8 Saving plots to files with ggsave()

You can save the last plot as a PDF:

# Save last displayed plot as PDF
ggsave("figs/myggplot.pdf")

Or save a named plot object:

p = ggplot(data = dat,
       mapping = aes(x = ReadingIRTScore,
                     y = MathIRTScore,
                     colour = EnrolmentStatus)) +
  geom_point(alpha = 0.25) +
  labs(
    x = "Reading latent ability score",
    y = "Maths latent ability score",
    colour = "School enrollment status",
    title = "Foundational Skills in Senegal",
    subtitle = "ICAN-ICAR 2025 Survey"
  ) +
  scale_x_continuous(labels = label_comma()) +
  scale_y_continuous(labels = label_comma()) +
  theme_classic()

ggsave(filename = "figs/myggplot.png",
       plot   = p,
       width  = 10,
       height = 10,
       units  = "in")
  • The file type is guessed from the extension (.pdf, .png, etc.).
  • You can control size with width, height, and units.

4.9 10. Useful resources

Some excellent follow-up resources:

# Friendly reminder: once you're comfortable with the basics,
# pick one plot you make often and try to recreate it using ggplot2.It can be a bar graph, histogram, or a box plot