Introductory Visualisation

RAdelaide 2025

Author

Affiliation

Dr Stevie Pederson

Black Ochre Data Labs
The Kids Research Institute Australia

Visualisation in R

Start a New R Script

• Call the new script: IntroVisualisation.R
• Load our favourite packages at the top of the script

library(palmerpenguins)
library(tidyverse)

Base Plotting in R

• R comes with some very powerful plotting capabilities
  • Provided in the base package graphics
  • Always loaded with every session
• Examples are often extremely helpful
• People used happily for decades
  • The release of ggplot2 changed everything
• Let’s quickly explore base plotting before moving to the good stuff

Base Plotting In R

• Simple plots are usually easy
  • Complex figures can get really messy
• Using the cars dataset
  • speed (mph)
  • dist (ft) each car takes to stop

plot(cars)

Base Plotting In R

• The first two columns were automatically placed on the x & y axis
• We could set values for x & y manually
  • Switching back to the penguins here
• Automatically decided to plot using points

## Plot calling individual columns from penguins using `$`
plot(x = penguins$bill_depth_mm, y = penguins$bill_length_mm)

Base Plotting In R

• The function boxplot() can also create simple figures easily
• For categorical variables (i.e. factors) we can use the formula notation
  • y ~ x \(\implies\) y depends on x

## Make a simple boxplot showing the weights by species
boxplot(body_mass_g ~ species, data = penguins)

• The dependent variable will always appear on the y-axis
• The predictor will always appear on the x-axis

Base Plotting In R

• We can also use combinations of predictor variables

## Separate by species and sex
boxplot(body_mass_g ~ sex + species, data = penguins)

Base Plotting In R

• Histograms can be produced on an individual column
  • The number of breaks can be set manually
• The default is pretty useful here
  • Generally simple figures without complexity

hist(penguins$body_mass_g, breaks = 20, xlab = "Body Mass (g)")

Base Plotting In R

• Large datasets can be quickly explored using pairs()
• Shows all pairwise combinations of columns
  • Categorical columns can be less informative

pairs(penguins)

• When we called plot on cars, this was actually called under the hood
• Only two columns to show in a pairwise manner

Visualisation With ggplot2

The Grammar of Graphics

• ggplot2 has become the industry standard for visualisation (Wickham 2016)
• Core & essential part of the tidyverse
• Developed by Hadley Wickham as his PhD thesis
• An implementation of The Grammar of Graphics (Wilkinson 2005)
  • Breaks visualisation into layers

The Grammar of Graphics

Taken from https://r.qcbs.ca/workshop03/book-en/grammar-of-graphics-gg-basics.html

The Grammar of Graphics

Everything is added in layers

1. Data
   • Usually a data.frame (or tibble)
   • Can be piped in \(\implies\) modify on the fly

2. Aesthetics
   • x & y co-ordinates
   • colour, fill, shape, size, linetype
   • grouping & transparency (alpha)

3. Geometric Objects
   • points, lines, boxplot, histogram, bars etc

4. Facets: Panels within plots

5. Statistics: Computed summaries

6. Coordinates
   • polar, map, cartesian etc
   • defaults to cartesian

7. Themes: overall layout
   • default themes automatically applied

An Initial Example

• Using the example dataset cars
• Two columns:
  • speed (mph)
  • distance each car takes to stop
• We can make a classic x vs y plot using points

• The predictor (x) would be speed
• The response (y) would be distance

An Initial Example

• We may as well start by piping our data in

cars |>
  ggplot(aes(x = speed, y = dist))

• We have defined the plotting aesthetics
  • x & y
  • Don’t need to name if passing in order
• Axis limits match the data

• No geometry has been specified \(\implies\) nothing was drawn

• The package is ggplot2 but the function is ggplot()

An Initial Example

• To add points, we add geom_point() after calling ggplot()
• Adding + after ggplot() says “But wait! There’s more…”

cars |>
  ggplot(aes(x = speed, y = dist)) + 
  geom_point() 

• When ggplot2 was created neither pipe had been developed yet

An Initial Example

• To add points, we add geom_point() after calling ggplot()
  • Adding + after ggplot() says “But wait! There’s more…”

cars |> # Layer 1: Data
  ggplot(aes(x = speed, y = dist)) + # Layer 2: Aesthetics
  geom_point() # Layer 3: Geometry

• By default:
  • Layer 4: No facets
  • Layer 5: No summary statistics
  • Layer 6: Cartesian co-ordinate system
  • Layer 7: Crappy theme with grey background 🤮

Axis limits are automatically determined

An Initial Example

• A simple summary statistic to add might be stat_smooth()
• Automatically chooses the smoother
  • Usually a loess curve or regression line
  • The standard error region is shown by default

cars |> # Layer 1: Data
  ggplot(aes(x = speed, y = dist)) + # Layer 2: Aesthetics
  geom_point() + # Layer 3: Geometry
  stat_smooth() # Layer 5: Statistics

Visualising Our Penguins

What visualisations could we produce to inspect penguins?

• Boxplots & Points

Creating Our First Plot

## Compare the two bill measurements
penguins |> # Layer 1: Data
  ggplot(aes(x = bill_depth_mm, y = bill_length_mm)) + # Layer 2: Aesthetics
  geom_point()  # Layer 3: Geometry

Creating Our First Plot

• There seem to be groups. Are these based on species? \(\implies\) Add colour

## Compare the two bill measurements
penguins |> 
  ggplot(aes(x = bill_depth_mm, y = bill_length_mm, colour = species)) +
  geom_point()

Creating Our First Plot

• We can also add regression lines
  • We’ll add equations later
  • Try without the se = FALSE and see what happens

## Add regression lines as a new geom
penguins |>
  ggplot(aes(x = bill_depth_mm, y = bill_length_mm, colour = species)) +
  geom_point() +
  stat_smooth(method = "lm", se = FALSE) # Layer 5: Statistics

Understanding Aesthetics

• Setting the colour in the call to ggplot() \(\implies\) all layers will use this
• If we shift colour = species to geom_point() \(\implies\) ???

## Only use colour for the points
penguins |>
  ggplot(aes(x = bill_depth_mm, y = bill_length_mm)) +
  geom_point(aes(colour = species)) +
  stat_smooth(method = "lm", se = FALSE)

Understanding Aesthetics

• We could set this again if we choose

## Set colour for the points and regression lines separately
penguins |>
  ggplot(aes(x = bill_depth_mm, y = bill_length_mm)) +
  geom_point(aes(colour = species)) +
  stat_smooth(aes(colour = species), method = "lm", se = FALSE)

• It’s clunky here, but can give fine control for complex plots

Using Facets

• Alternatively, we could plot each species in it’s own panel (or facet)
• Using ~ notation to say all facets depend on species

## Plot each species in a separate panel
penguins |>
  ggplot(aes(x = bill_depth_mm, y = bill_length_mm)) +
  geom_point(aes(colour = species)) +
  stat_smooth(method = "lm", se = FALSE) +
  facet_wrap(~species) # Layer 4: Facets

Using Facets

• We can allow x and y axes to scale separately for each panel
  • Not always a helpful strategy

## Plot each species in a separate panel, allowing axes to be scaled freely
penguins |>
  ggplot(aes(x = bill_depth_mm, y = bill_length_mm)) +
  geom_point(aes(colour = species)) +
  stat_smooth(method = "lm", se = FALSE) +
  facet_wrap(~species, scales = "free") 

Scales

Setting Scales

• By default, ggplot2 will detect the most appropriate scale
  • Has applied scale_x_continuous() and scale_y_continuous()

# Explicitly set the scales. This will appear identical
penguins |>
  ggplot(aes(x = bill_depth_mm, y = bill_length_mm, colour = species)) +
  geom_point() +
  stat_smooth(method = "lm", se = FALSE) +
  scale_x_continuous() +
  scale_y_continuous()

• Multiple presets are available:
  • scale_x_log10(), scale_x_sqrt(), scale_x_reverse()
  • Also available for y

Setting Scales

• For aesthetics like colour, we often want to tailor these
  • Default is scale_colour_discrete() (Meh…)
• Many defaults exist
  • scale_colour_brewer(), scale_colour_viridis_d()

## Check the default palette for scale_colour_brewer
penguins |>
  ggplot(aes(x = bill_depth_mm, y = bill_length_mm, colour = species)) +
  geom_point() +
  stat_smooth(method = "lm", se = FALSE) +
  scale_colour_brewer()

• Default palettes can be good sometimes
  • To show options for scale_colour_brewer() \(\implies\) RColorBrewer::display.brewer.all()

Setting Scales

• I often use Set1, but try a few others
• scale_colour_viridis_d() will give a colourblind-friendly palette
  • Other palettes are provided by other packages

## Set the palette for scale_colour_brewer to be "Set1" or anything else
penguins |>
  ggplot(aes(x = bill_depth_mm, y = bill_length_mm, colour = species)) +
  geom_point() +
  stat_smooth(method = "lm", se = FALSE) +
  scale_colour_brewer(palette = "Set1")

Setting Scales

• Standard 7-colour palette adapted for colourblindness is included in ggthemes (Wong 2011)
  • Many alternatives exist
  • This one is written by Americans \(\implies\) weird spelling of colourblind

library(ggthemes)
## Use the colourblind friendly palette provided by ggthemes
penguins |>
  ggplot(aes(x = bill_depth_mm, y = bill_length_mm, colour = species)) +
  geom_point() +
  stat_smooth(method = "lm", se = FALSE) +
  scale_colour_colorblind()

• The colourblind-friendly status can be checked using RGB Hex codes at https://davidmathlogic.com/colorblind/#%23D81B60-%231E88E5-%23FFC107-%23004D40

What Else Can We Do?

• What else might be informative?

• Can we separate by island or sex?
  • sex will have missing values
  • Let’s set the shape of the points

## Try setting different point shapes based on the recorded sex
penguins |>
  filter(!is.na(sex)) |> # Remove the penguins with unrecorded sex
  ggplot(aes(x = bill_depth_mm, y = bill_length_mm, colour = species)) +
  geom_point(aes(shape = sex)) + # Now we have a layer-specific aesthetic 
  stat_smooth(method = "lm", se = FALSE) +
  scale_colour_colorblind()

Modifying Points

• We can change the size of these outside the aesthetic
  • Fixed values only \(\implies\) will not respond to change in data

## Try setting different point shapes based on the recorded sex
penguins |>
  filter(!is.na(sex)) |> # Remove the penguins with unrecorded sex
  ggplot(aes(x = bill_depth_mm, y = bill_length_mm, colour = species)) +
  geom_point(aes(shape = sex), size = 3) + # Change the point size
  stat_smooth(method = "lm", se = FALSE) +
  scale_colour_colorblind()

Modifying Points

• Points can be set manually using scale_shape_manual()
  • Also scale_colour_manual()

## Try setting different point shapes based on the recorded sex
penguins |>
  filter(!is.na(sex)) |> # Remove the penguins with unrecorded sex
  ggplot(aes(x = bill_depth_mm, y = bill_length_mm, colour = species)) +
  geom_point(aes(shape = sex), size = 3) + 
  stat_smooth(method = "lm", se = FALSE) +
  scale_colour_colorblind() +
  scale_shape_manual(values = c(19, 1)) ## Manually choose the point shapes

Modifying Points

• How did I know to choose those two values?
• Why do numbers represent different shapes

• Enter ?pch and scroll down a little
  • 21-25 have both a colour (outline) and fill capability

Finishing Our Figure

• The next step in making our figure look brilliant
  • Axis & Scale labels

## Try setting different point shapes based on the recorded sex
penguins |>
  filter(!is.na(sex)) |> # Remove the penguins with unrecorded sex
  ggplot(aes(x = bill_depth_mm, y = bill_length_mm, colour = species)) +
  geom_point(aes(shape = sex), size = 3) + 
  stat_smooth(method = "lm", se = FALSE) +
  scale_colour_colorblind() +
  scale_shape_manual(values = c(19, 1)) +
  labs(
    # Manually add labels
    x = "Bill Depth (mm)", y = "Bill Length (mm)", 
    colour = "Species", shape = "Sex"
  ) 

• Note that this fundamentally breaks the automatic association between our data & figure
• We can type literally anything for labels which can leave to door open to errors

Finishing Our Figure

• The final layer in the Grammar of Graphics is the Theme
• Controls the overall appearance not controlled elsewhere
• The code can get long so let’s save that figure as p
  • Then I can modify on a single slide

## Save the figure for exploring theme attributes
p <- penguins |>
  filter(!is.na(sex)) |> # Remove the penguins with unrecorded sex
  ggplot(aes(x = bill_depth_mm, y = bill_length_mm, colour = species)) +
  geom_point(aes(shape = sex), size = 3) + 
  stat_smooth(method = "lm", se = FALSE) +
  scale_colour_colorblind() +
  scale_shape_manual(values = c(19, 1)) +
  labs(
    x = "Bill Depth (mm)", y = "Bill Length (mm)", 
    colour = "Species", shape = "Sex"
  ) 

Themes

Using Themes

• A default theme is applied: theme_grey()
• I prefer theme_bw()
  • Removes the grey background
  • Gets most of the job done

p + theme_bw()

Using Themes

• Additional modifications:
  • Setting the base font size for all annotations
  • Can also set colour, alignment, font face, font family etc
  • Vital for publishing figures

p + theme_bw() +
  theme(text = element_text(size = 14))

Using Themes

• A key application is the placement of legends

p + theme_bw() +
  theme(legend.position = "bottom")

Using Themes

• Legend can be placed inside using three steps
  • Set legend.position = "inside"
  • Set the position you want the legend
  • Set which part of the legend aligns at those co-ordinates
• Can be extremely finicky

p + theme_bw() +
  theme(
    legend.position = "inside", # Ensure the legend is inside the plotting region
    legend.position.inside = c(0, 0), # Anchor to the bottom left
    legend.justification.inside = c(0, 0) # Set the alignment to be bottom left
  )

• By default the legend will align the centre at the given co-ordinates

Using Themes

• Any individual text element of a figure can be modified using element_text()

p + theme_bw() +
  theme(
    ## A slightly exaggerated modification of axis titles
    axis.title = element_text(colour = "darkred", size = 16, face = "bold")
  )

Using Themes

• Looking at the help page ?theme \(\implies\) 4 main types of ‘element’

1. element_text()
   • Control all text elements (axes, titles etc)
   • Doesn’t impact labels within figures

2. element_line()
   • Control all lines (axes, gridlines, borders etc)
   • Can set colour, size, linetype, linewidth

3. element_rect()
   • Control all rectangles (background, legends, panels etc)
   • Can set colour, fill, size, linetype, linewidth

4. element_blank()
   • Hides an element

• Other element types are a bit more nuanced

Using Themes

# Make the most horrible figure possible
p + theme_bw() +
  theme(
    ## A slightly exaggerated modification of axis titles
    axis.title = element_text(colour = "darkred", size = 16, face = "bold"),
    ## Make axes thick, blue lines. Ewww
    axis.line = element_line(colour = "darkblue", linewidth = 2),
    ## Hide the underlying grid
    panel.grid = element_blank(),
    ## Make the area background a light grey
    plot.background = element_rect(fill = "grey70")
  )

Using Themes

• Plot titles align left by default

p + theme_bw() + 
  ggtitle("Penguin Bill Measurements")

• We can use theme() to align in the centre
  • hjust is the horizontal adjustment

p + theme_bw() + 
  ggtitle("Penguin Bill Measurements") +
  theme(plot.title = element_text(hjust = 0.5))

Different Plot Types

Different Plot Types

Classic BarPlots

• geom_bar() & geom_col()
• geom_errorbar() & geom_errorbarh()

Classic Density plots

• geom_boxplot() & geom_violin()
• geom_density() & geom_histogram()

Line-based Geometry

• geom_line(), geom_segment()
• geom_abline(), geom_hline() & geom_vline()

Heatmaps and Grids

• geom_raster(), geom_tile() & geom_rect()

Creating A Boxplot

• A starting point might be to choose sex as the predictor
• body_mass_g may be a response variable

penguins |> 
  ggplot(aes(island, body_mass_g)) +
  geom_boxplot()

Creating Our Boxplot

• To incorporate the sex \(\implies\) add a fill aesthetic
  • colour is generally applied to shape outlines

## Fill the boxes by sex
penguins |> 
  ggplot(aes(island, body_mass_g, fill = sex)) +
  geom_boxplot()

• ggplot2 will always separate multiple values/category

Creating Our Boxplot

## Remove the penguins with no recorded sex
penguins |> 
  filter(!is.na(sex)) |>
  ggplot(aes(island, body_mass_g, fill = sex)) +
  geom_boxplot()

Creating Our Boxplot

• We could also separate by island using facet_wrap()

penguins |> 
  filter(!is.na(sex)) |>
  ggplot(aes(island, body_mass_g, fill = sex)) +
  geom_boxplot() +
  facet_wrap(~species, scales = "free_x")

Creating Our Boxplot

• A less-intuitive alternative (facet_grid()) will allow for unequal-sized facets

penguins |> 
  filter(!is.na(sex)) |>
  ggplot(aes(island, body_mass_g, fill = sex)) +
  geom_boxplot() +
  facet_grid(~species, scales = "free_x", space = "free_x")

Overlaying Geoms

• geom_jitter() will draw points but with noise in either direction
• The alpha parameter will make the points partially transparent

penguins |> 
  filter(!is.na(sex)) |>
  ggplot(aes(sex, body_mass_g, fill = sex)) +
  geom_boxplot(outliers = FALSE) + # Hide any outliers from the boxes
  geom_jitter(width = 0.1, alpha = 0.5) + # Outliers will be shown here
  facet_wrap(~species + island, nrow = 1)

Trying a Violin Plot

• Violin plots are similar to boxplots
  • can estimate distributions beyond the points (trim = FALSE)

penguins |> 
  filter(!is.na(sex)) |>
  ggplot(aes(sex, body_mass_g, fill = sex)) +
  geom_violin(draw_quantiles = 0.5, trim = FALSE) +
  geom_jitter(width = 0.1, alpha = 0.5) +
  facet_wrap(~species + island, nrow = 1)

Creating A Histogram

• The default histogram in ggplot2 is a bit ugly

penguins |>
  filter(!is.na(sex)) |>
  ggplot(aes(body_mass_g)) +
  geom_histogram()

• This can be easily improved by setting fill = "grey70" and colour = "black"
• binwidth is automatically set \(\implies\) try binwidth = 100

Creating A Histogram

## Now apply facet_grid to show by sex & species
penguins |>
  filter(!is.na(sex)) |>
  ggplot(aes(body_mass_g)) +
  geom_histogram(fill = "grey70", colour = "black", binwidth = 100) +
  facet_grid(species ~ sex) +
  theme_bw()

Creating A Histogram

• The above shows counts \(\implies\) can also show frequency (i.e. density)
• Unfortunately, the code is ugly but does make sense
  • (The ‘stat’ is counting, so after counting convert to a frequency)

## Repeat but showing bars as frequencies
penguins |>
  filter(!is.na(sex)) |>
  ggplot(aes(body_mass_g)) +
  geom_histogram(
    aes(y = after_stat(density)),
    fill = "grey70", colour = "black", binwidth = 100
  ) +
  facet_grid(species ~ sex) +
  theme_bw()

• The first plot is informative if we have roughly the same number in each panel
• The second option can be more informative for unbalanced data

Creating a Summary Barplot

• Bar plots are pretty common in many fields of research
• First we’ll create a summary table with mean and sd
  • Use the mean for bars
  • sd for error bars

penguins |>
  filter(!is.na(sex)) |>
  summarise(
    weight_mn = mean(body_mass_g, na.rm = TRUE),
    weight_sd = sd(body_mass_g, na.rm = TRUE),
    .by = c(species, sex)
  ) 

# A tibble: 6 × 4
  species   sex    weight_mn weight_sd
  <fct>     <fct>      <dbl>     <dbl>
1 Adelie    male       4043.      347.
2 Adelie    female     3369.      269.
3 Gentoo    female     4680.      282.
4 Gentoo    male       5485.      313.
5 Chinstrap female     3527.      285.
6 Chinstrap male       3939.      362.

Creating a Summary Barplot

• Now we can use geom_col()

## Begin creating a bar plot, with separate panels for each species
penguins |>
  filter(!is.na(sex)) |>
  summarise(
    weight_mn = mean(body_mass_g, na.rm = TRUE),
    weight_sd = sd(body_mass_g, na.rm = TRUE),
    .by = c(species, sex)
  ) |>
  ggplot(aes(sex, weight_mn, fill = sex)) +
  geom_col() +
  facet_wrap(~species, nrow = 1) 

Creating a Summary Barplot

penguins |>
  filter(!is.na(sex)) |>
  summarise(
    weight_mn = mean(body_mass_g, na.rm = TRUE),
    weight_sd = sd(body_mass_g, na.rm = TRUE),
    .by = c(species, sex)
  ) |>
  ggplot(aes(sex, weight_mn, fill = sex)) +
  geom_col() +
  ## Now add error bars adding/subtracting from the mean 'on the fly'
  geom_errorbar(
    aes(ymin = weight_mn - weight_sd, ymax = weight_mn + weight_sd),
    width = 0.2 # Set the width of tails on the error bars
  ) +
  facet_wrap(~species, nrow = 1) +
  ## Some extra code to make the plot look great
  scale_y_continuous(expand = expansion(c(0, 0.05))) +
  scale_fill_brewer(palette = "Set1") +
  theme_bw()

Adding Labels To Points

• The penguins dataset isn’t well suited to adding labels
  • Let’s make a toy dataset

tibble(
  x = 1:4, y = x^2, label = c("a", "b", "c", "d")
)

# A tibble: 4 × 3
      x     y label
  <int> <dbl> <chr>
1     1     1 a    
2     2     4 b    
3     3     9 c    
4     4    16 d    

Adding Labels To Points

• We can easily plot points, but what about labels?

tibble(
  x = 1:4, y = x^2, label = c("a", "b", "c", "d")
) |> 
  ggplot(aes(x, y)) +
  geom_point()

Adding Labels To Points

• Using geom_text() will overlay labels exactly on the points

tibble(
  x = 1:4, y = x^2, label = c("a", "b", "c", "d")
) |> 
  ggplot(aes(x, y)) +
  geom_point() +
  geom_text(aes(label = label), size = 4)

Adding Labels To Points

• Using geom_text_repel() will shift labels marginally away from the points

library(ggrepel)
tibble(
  x = 1:4, y = x^2, label = c("a", "b", "c", "d")
) |> 
  ggplot(aes(x, y)) +
  geom_point() +
  geom_text_repel(aes(label = label), size = 4)

• geom_label() and geom_label_repel() will add borders and fill to labels

Saving Images

• The simple way is click Export in the Plots pane

• The way to save using code is

ggsave("myplot.png", width = 7, height = 7, units = "in")

• This will always save the most recent plot by default
• Output format is determined by the suffix
• Try saving as a pdf…

Saving Images

• I think saving using code is preferable
• Modify an analysis or data \(\implies\) saved figures will also update
  • This saves time & ensures reproducibility
  • Modifying font sizes etc for publication can take a while

Conclusion

Conclusion

A fabulous resource: https://r-graphics.org/

Challenges

1. Create a barplot with error bars showing mean flipper length by species
   • Colour (i.e. fill) however you choose
2. Create a histogram of flipper length by species
   • Facet by species
3. Use boxplots to show the same (flipper length by species)
   • Don’t facet, but fill the boxes by sex
4. Using points, compare flipper length to body mass
   • Colour by species including a regression line
   • Try adding stat_ellipse() to your plot

References

Wickham, Hadley. 2016. Ggplot2: Elegant Graphics for Data Analysis. Springer-Verlag New York. https://ggplot2.tidyverse.org.

Wilkinson, Leland. 2005. The Grammar of Graphics. Springer New York, NY. https://doi.org/https://doi.org/10.1007/0-387-28695-0.

Wong, Bang. 2011. “Color Blindness.” Nat. Methods 8 (6): 441.