Workshop: Dealing with Data in R

Visualizing Data in R

A primer on ggplot2

steffilazerte
@steffilazerte@fosstodon.org
@steffilazerte
steffilazerte.ca

Compiled: 2024-02-21

First things first

Save previous script

Open New File
(make sure you’re in the RStudio Project)

Write library(tidyverse) at the top

Save this new script
(consider names like figs.R or 2_figures.R)

Outline

1. Figures with ggplot2 (A tidyverse package)

• Basic plot
• Common plot types
• Plotting by categories
• Adding statistics
• Customizing plots
• Annotating plots

2. Combining figures with patchwork

3. Saving figures

Artwork by @allison_horst

Our data set: Palmer Penguins!

Artwork by @allison_horst

Our data set: Palmer Penguins!

library(palmerpenguins)
penguins

# A tibble: 344 × 8
   species island    bill_length_mm bill_depth_mm flipper_length_mm body_mass_g
   <fct>   <fct>              <dbl>         <dbl>             <int>       <int>
 1 Adelie  Torgersen           39.1          18.7               181        3750
 2 Adelie  Torgersen           39.5          17.4               186        3800
 3 Adelie  Torgersen           40.3          18                 195        3250
 4 Adelie  Torgersen           NA            NA                  NA          NA
 5 Adelie  Torgersen           36.7          19.3               193        3450
 6 Adelie  Torgersen           39.3          20.6               190        3650
 7 Adelie  Torgersen           38.9          17.8               181        3625
 8 Adelie  Torgersen           39.2          19.6               195        4675
 9 Adelie  Torgersen           34.1          18.1               193        3475
10 Adelie  Torgersen           42            20.2               190        4250
# ℹ 334 more rows
# ℹ 2 more variables: sex <fct>, year <int>

Artwork by @allison_horst

Your turn!

Run this code and look at the output in the console

Side Note

Where did the penguins data set come from?

• Sometimes R packages contain data
• If you load a package (i.e. library(palmerpenguins)) you can use the data
• Note that here the data object is called penguins (not palmerpenguins)
• Note this is NOT how you’ll load your own data

A basic plot

library(palmerpenguins)
library(tidyverse)

ggplot(data = penguins, aes(x = body_mass_g, y = bill_length_mm)) +
    geom_point()

Break it down

library(palmerpenguins)
library(tidyverse)

ggplot(data = penguins, aes(x = body_mass_g, y = bill_length_mm)) +
    geom_point()

library()

• Load the palmerguins package
• Now we have access to penguins data

Break it down

library(palmerpenguins)
library(tidyverse)

ggplot(data = penguins, aes(x = body_mass_g, y = bill_length_mm)) +
    geom_point()

library()

• Load the tidyverse packages
  (includes ggplot2)
• Now we have access to the ggplot() function (and aes() and geom_point() etc.)

Break it down

library(palmerpenguins)
library(tidyverse)

ggplot(data = penguins, aes(x = body_mass_g, y = bill_length_mm)) +
    geom_point()

ggplot()

• Set the attributes of your plot
• data = Dataset
• aes = Aesthetics (how the data are used)
• Think of this as your plot defaults

Break it down

library(palmerpenguins)
library(tidyverse)

ggplot(data = penguins, aes(x = body_mass_g, y = bill_length_mm)) +
    geom_point()

geom_point()

• Choose a geom function to display the data
• Always added to a ggplot() call with +

ggplots are essentially layered objects, starting with a call to ggplot()

Plots are layered

ggplot(data = penguins, aes(x = sex, y = body_mass_g))

ggplot(data = penguins, aes(x = sex, y = body_mass_g)) +
  geom_boxplot()

ggplot(data = penguins, aes(x = sex, y = body_mass_g)) +
  geom_point()

ggplot(data = penguins, aes(x = sex, y = body_mass_g)) +
  geom_violin()

Plots are layered

You can add multiple layers

ggplot(data = penguins, aes(x = sex, y = body_mass_g)) +
  geom_boxplot()

ggplot(data = penguins, aes(x = sex, y = body_mass_g)) +
  geom_boxplot() +
  geom_point(size = 2, colour = "red", 
             position = position_jitter(width = 0.05))

Order matters

ggplot(data = penguins, aes(x = sex, y = body_mass_g)) +
  geom_point(size = 2, colour = "red",
             position = position_jitter(width = 0.05)) +
  geom_boxplot()

Plots are objects

Any ggplot can be saved as an object

g <- ggplot(data = penguins, aes(x = sex, y = body_mass_g))

g

g + geom_boxplot()

More Geoms

(Plot types)

Geoms: Lines

ggplot(data = penguins, aes(x = body_mass_g, y = bill_length_mm)) +
  geom_line()

Geoms: Boxplots

ggplot(data = penguins, aes(x = species, y = flipper_length_mm)) +
  geom_boxplot() 

Geoms: Histogram

ggplot(data = penguins, aes(x = body_mass_g)) +
  geom_histogram(binwidth = 100)

Note:
We only need 1 aesthetic here

Geoms: Barplots

Let ggplot count your data

ggplot(data = penguins, aes(x = sex)) +
  geom_bar()

Geoms: Barplots

You can also provide the counts

# Create our own data frame
species_counts <- data.frame(species = c("Adelie", "Chinstrap", "Gentoo"),
                             n = c(152, 68, 124))

ggplot(data = species_counts, aes(x = species, y = n)) +
  geom_bar(stat = "identity")

Your Turn: Create this plot

ggplot(data = ____, aes(x = ____, y = ____)) +
  geom_____(____)

Too Easy?
Plot points on top
Why not consider jittering them?

Your Turn: Create this plot

ggplot(data = penguins, aes(x = island, y = bill_depth_mm)) +
  geom_boxplot(colour = "blue")

Your Turn: Create this plot

Too Easy?

ggplot(data = penguins, aes(x = island, y = bill_depth_mm)) +
  geom_boxplot(colour = "blue") +
  geom_point()

Your Turn: Create this plot

Too Easy?

ggplot(data = penguins, aes(x = island, y = bill_depth_mm)) +
  geom_boxplot(colour = "blue") +
  geom_point(position = position_jitter(width = 0.1))

Your Turn: Create this plot

Too Easy?

ggplot(data = penguins, aes(x = island, y = bill_depth_mm)) +
  geom_boxplot(colour = "blue") +
  geom_count()

Showing data by group

Mapping aesthetics

ggplot(data = penguins, aes(x = body_mass_g, y = bill_length_mm)) +
  geom_point()

Mapping aesthetics

ggplot(data = penguins, aes(x = body_mass_g, y = bill_length_mm, colour = sex)) +
  geom_point()

Mapping aesthetics

ggplot automatically populates the legends (combining where it can)

ggplot(data = penguins, aes(x = body_mass_g, y = bill_length_mm, colour = sex, shape = sex)) +
  geom_point()

Faceting: facet_wrap()

ggplot(data = penguins, aes(x = body_mass_g, y = bill_length_mm, colour = sex)) +
  geom_point() +
  facet_wrap(~ species)

Split plots by one grouping variable

Faceting: facet_grid()

ggplot(data = penguins, aes(x = body_mass_g, y = bill_length_mm, colour = sex)) +
  geom_point() +
  facet_grid(sex ~ species)

Split plots by two grouping variables

Your Turn: Create this plot

ggplot(data = ____, aes(_____________________________________)) +
  ______________ +
  ______________

Hint: colour is for outlining with a colour, fill is for ‘filling’ with a colour
Too Easy? Split boxplots by sex and island

Your Turn: Create this plot

ggplot(data = penguins, aes(x = sex, y = flipper_length_mm, fill = sex)) +
  geom_boxplot() +
  facet_wrap(~ species)

Hint: colour is for outlining with a colour, fill is for ‘filling’ with a colour
Too Easy? Split boxplots by sex and island

Your Turn: Create this plot

Too Easy?

ggplot(data = penguins, aes(x = sex, y = flipper_length_mm, fill = island)) +
  geom_boxplot() +
  facet_wrap(~ species)

Small change (fill = sex to fill = island) results in completely different plot

Adding Statistics to Plots

Summarizing data

Add data means as points

ggplot(data = penguins, aes(x = sex, y = body_mass_g)) +
  stat_summary(geom = "point", fun = mean)

Summarizing data

Add error bars, calculated from the data

ggplot(data = penguins, aes(x = sex, y = body_mass_g)) +
  stat_summary(geom = "point", fun = mean) +
  stat_summary(geom = "errorbar", width = 0.05, fun.data = mean_se)

Trendlines / Regression Lines

Trendlines / Regression lines

geom_line() is connect-the-dots, not a trend or linear model

ggplot(data = penguins, aes(x = body_mass_g, y = bill_length_mm)) +
  geom_point() +
  geom_line()

Not what we’re looking for

Trendlines / Regression lines

Let’s add a trend line properly

Start with basic plot:

g <- ggplot(data = penguins, aes(x = body_mass_g, y = bill_length_mm)) +
  geom_point()
g

Trendlines / Regression lines

Add the stat_smooth()

• lm is for “linear model” (i.e. trendline)
• grey ribbon = standard error

g + stat_smooth(method = "lm")

Trendlines / Regression lines

Add the stat_smooth()

• remove the grey ribbon se = FALSE

g + stat_smooth(method = "lm", se = FALSE)

Trendlines / Regression lines

A line for each group

• Specify group (here we use colour to specify species)

g <- ggplot(data = penguins, aes(x = body_mass_g, y = bill_length_mm, colour = species)) +
  geom_point()
g

Trendlines / Regression lines

A line for each group

• stat_smooth() automatically uses the same grouping

g + stat_smooth(method = "lm", se = FALSE)

Trendlines / Regression lines

A line for each group AND overall

g +
  stat_smooth(method = "lm", se = FALSE) +
  stat_smooth(method = "lm", se = FALSE, colour = "black")

Your Turn: Create this plot

• A scatter plot: Flipper Length by Body Mass grouped by Species
• With a single regression line for the overall trend

Too Easy? Add regression lines for each species as well
Can you make the species lines larger?
Can you indicate which points are female and which are male?

Your Turn: Create this plot

• A scatter plot: Flipper Length by Body Mass grouped by Species
• With a single regression line for the overall trend

ggplot(data = penguins, aes(x = body_mass_g, y = flipper_length_mm, colour = species)) +
  geom_point() +
  stat_smooth(method = "lm", se = FALSE, colour = "black")

Your Turn: Create this plot

Too Easy?

ggplot(data = penguins, aes(x = body_mass_g, y = flipper_length_mm, colour = species)) +
  geom_point() +
  stat_smooth(method = "lm", se = FALSE) +
  stat_smooth(method = "lm", se = FALSE, colour = "black")

Your Turn: Create this plot

Too Easy?

ggplot(data = penguins, aes(x = body_mass_g, y = flipper_length_mm, 
                            colour = species)) +
  geom_point(aes(shape = sex), size = 2, fill = "white") +
  stat_smooth(method = "lm", se = FALSE, linewidth = 2) +
  stat_smooth(method = "lm", se = FALSE, colour = "black") +
  scale_shape_manual(values = c(20, 21))

Customizing plots

Customizing: Starting plot

Let’s work with this plot

g <- ggplot(data = penguins, aes(x = body_mass_g, y = bill_length_mm, colour = species)) +
  geom_point()

Customizing: Labels

g + labs(title = "Bill Length vs. Body Mass",
         x = "Body Mass (g)",
         y = "Bill Length (mm)",
         colour = "Species", tag = "A")

Your Turn: Add proper labels to some of your previous plots

Customizing: Built-in themes

Customizing: Axes

scale_ + (x or y) + type (continuous, discrete, date, datetime)

• scale_x_continuous()
• scale_y_discrete()
• etc.

Common arguments

g + scale_x_continuous(breaks = seq(0, 20, 10)) # Tick breaks
g + scale_x_continuous(limits = c(0, 15))       # xlim() is a shortcut for this
g + scale_x_continuous(expand = c(0, 0))        # Space between axis and data

Let’s take a look…

Customizing: Axes

Breaks

g + scale_x_continuous(breaks = seq(2500, 6500, 500))

Customizing: Axes

Limits

g + scale_x_continuous(limits = c(3000, 4000))

Customizing: Axes

Space between origin and axis start

g + scale_x_continuous(expand = c(0, 0))

Customizing: Aesthetics

Using scales

scale_ + aesthetic (colour, fill, size, etc.) + type (manual, continuous, datetime, etc.)

g + scale_colour_manual(name = "Type", values = c("green4", "blue4", "gold"))

Customizing: Aesthetics

Using scales

Or be very explicit:

g + scale_colour_manual(
  name = "Type",
  values = c("Adelie" = "green4", "Gentoo" = "blue4", "Chinstrap" = "gold"),
  na.value = "black")

Customizing: Aesthetics

For colours, consider colour-blind-friendly scale

viridis_d for “discrete” data

ggplot(data = penguins, aes(x = body_mass_g, y = bill_length_mm, colour = species)) +
  geom_point() +
  scale_colour_viridis_d(name = "Type")

Customizing: Aesthetics

For colours, consider colour-blind-friendly scale

viridis_c for “continuous” data

ggplot(data = penguins, aes(x = body_mass_g, y = bill_length_mm, colour = flipper_length_mm)) +
  geom_point() +
  scale_colour_viridis_c(name = "Flipper Length (mm)")

Customizing: Aesthetics

Forcing

Remove the association between a variable and an aesthetic

ggplot(data = penguins, aes(x = body_mass_g, y = bill_length_mm, colour = sex)) +
  geom_point(colour = "darkblue", size = 1) +
  stat_smooth(method = "lm", se = FALSE, colour = "lightblue")

Note: When forcing, aesthetic is not inside aes()

Customizing: Legends placement

At the: top, bottom, left, right

g + theme(legend.position = "top")

Exactly here

g + theme(legend.position = c(0.15, 0.7))

Your Turn: Create this plot

Too Easy?
Play with shape values >20 and fill and colour
Or, create a plot of your own data

Your Turn: Create this plot

ggplot(penguins, aes(x = body_mass_g, y = flipper_length_mm, colour = species)) +
  theme_bw() + 
  geom_point() +
  stat_smooth(method = "lm", se = FALSE, colour = "black") + 
  scale_colour_viridis_d() +
  facet_wrap(~ sex) +
  labs(x = "Body Mass (g)",
       y = "Flipper Length (mm)",
       colour = "Species")

Your Turn: Create this plot

Too easy?

ggplot(penguins, aes(x = body_mass_g, y = flipper_length_mm, fill = species)) +
  theme_bw() + 
  geom_point(shape = 21) +
  stat_smooth(method = "lm", se = FALSE, colour = "black", fill = NA) + 
  scale_fill_viridis_d() +
  facet_wrap(~ sex) +
  labs(x = "Body Mass (g)",
       y = "Flipper Length (mm)",
       colour = "Species")

Side note: Order of operations

Order of operations

Remember…

• ggplot() is the default line (all options passed down)
• The other lines are added with the + (options only apply to this line)

Order of operations

Where to put the aes()?

Sometimes it doesn’t matter…

ggplot(penguins, aes(x = body_mass_g, 
                     y = flipper_length_mm, 
                     colour = species)) +
  geom_point()

ggplot(penguins, aes(x = body_mass_g, 
                     y = flipper_length_mm)) +
  geom_point(aes(colour = species))

Order of operations

Where to put the aes()?

Sometimes it DOES matter…

ggplot(penguins, aes(x = body_mass_g, 
                     y = flipper_length_mm, 
                     colour = species)) +
  geom_point() +
  stat_smooth(method = "lm")

ggplot(penguins, aes(x = body_mass_g, 
                     y = flipper_length_mm)) +
  geom_point(aes(colour = species)) +
  stat_smooth(method = "lm")

Applies to ALL lines in the ggplot
including stat_smooth()

Applies to only the geom_point() in the ggplot
not stat_smooth()

Combining plots with patchwork

Artwork by @allison_horst

Combining plots

Setup

• Load patchwork
• Create a couple of different plots

library(patchwork)

g1 <- ggplot(data = penguins, aes(x = bill_length_mm, y = bill_depth_mm, colour = species)) +
  geom_point()

g2 <- ggplot(data = penguins, aes(x = species, y = flipper_length_mm)) +
  geom_boxplot()

g3 <- ggplot(data = penguins, aes(x = flipper_length_mm, y = body_mass_g, colour = species)) +
  geom_point()

Combining plots with patchwork

Side-by-Side 2 plots

g1 + g2

Combining plots with patchwork

Side-by-Side 3 plots

g1 + g2 + g3

Combining plots with patchwork

Stacked 2 plots

g1 / g2

Combining plots with patchwork

More complex arrangements

g2 + (g1 / g3)

Combining plots with patchwork

More complex arrangements

g2 / (g1 + g3)

Combining plots with patchwork

“collect” common legends

g2 / (g1 + g3) + plot_layout(guides = "collect")

Combining plots with patchwork

“collect” common legends

g2 / (g1 + g3 + plot_layout(guides = "collect"))

Combining plots with patchwork

Annotate

g2 / (g1 + g3) +
  plot_layout(guides = "collect") +
  plot_annotation(title = "Penguins Data Summary",
                  caption = "Fig 1. Penguins Data Summary",
                  tag_levels = "A",
                  tag_suffix = ")")

Your Turn
Combine any 3 figures Too Easy?
Can you figure out how to collect common axes as well?

Your Turn: Combine plots

Too easy?

g1 <- ggplot(data = penguins, aes(x = bill_length_mm, y = bill_depth_mm, colour = species)) +
  geom_point()

g2 <- ggplot(data = penguins, aes(x = flipper_length_mm, y = bill_depth_mm, colour = species)) +
  geom_point()

g1 + g2 + plot_layout(guides = "collect", axes = "collect")

Saving plots

Saving plots

RStudio Export

Demo

ggsave()

g <- ggplot(penguins, aes(x = sex, y = bill_length_mm)) +
  geom_boxplot()

ggsave(filename = "penguins_mass.png", plot = g)

Saving plots

Publication quality plots

• Many publications require ‘lossless’ (pdf, svg, eps, ps) or high quality formats (tiff, png)
• Specific sizes corresponding to columns widths
• Minimum resolutions

g <- ggplot(penguins, aes(x = sex, y = body_mass_g)) +
  geom_boxplot() +
  labs(x = "Sex", y = "Body Mass (g)") +
  theme(axis.text.x = element_text(angle = 45, hjust = 1))

ggsave(filename = "penguins_mass.pdf", plot = g, dpi = 300,
       height = 80, width = 129, units = "mm")

Wrapping up: Common mistakes

• The package is ggplot2, the function is just ggplot()
• Did you remember to put the + at the end of the line?
• Order matters!
  • If you’re using custom theme()’s, make sure you put these lines after bundled themes like theme_bw(), or they will be overwritten
• Variables like ‘year’ are treated as continuous, but are really categories
  • Wrap them in factor()
  • e.g. ggplot(data = penguins, aes(x = factor(year), y = body_mass_g))

Wrapping up: Common mistakes

I get an error regarding an object that can’t be found or aesthetic length?

You are probably trying to plot two different datasets, and you make references to variables in the ggplot() call that don’t exist in one of the datasets:

n <- count(penguins, island)

ggplot(data = penguins, aes(x = flipper_length_mm, y = bill_length_mm, colour = species)) +
  geom_point() +
  facet_wrap(~ island) +
  geom_text(data = n, aes(label = n), 
            x = -Inf, y = +Inf, hjust = 0, vjust = 1)

Error in `geom_text()`:
! Problem while computing aesthetics.
ℹ Error occurred in the 2nd layer.
Caused by error:
! object 'species' not found

Wrapping up: Common mistakes

I get an error regarding an object that can’t be found or aesthetic length?

Either move the aesthetic…

ggplot(penguins, aes(x = flipper_length_mm, y = bill_length_mm)) +  
  geom_point(aes(colour = species)) +
  facet_wrap(~ island) +
  geom_text(data = n, aes(label = n), 
            x = -Inf, y = +Inf, hjust = 0, vjust = 1)

Wrapping up: Common mistakes

I get an error regarding an object that can’t be found or aesthetic length?

Either move the aesthetic…

Or assign it to NULL where it is missing…

ggplot(penguins, aes(x = flipper_length_mm, y = bill_length_mm, colour = species)) +  
  geom_point() +
  facet_wrap(~ island) +
  geom_text(data = n, aes(label = n, colour = NULL), 
            x = -Inf, y = +Inf, hjust = 0, vjust = 1)

Wrapping up: Further reading (all Free!)

• RStudio > Help > Cheatsheets > Data Visualization with ggplot2
• ggplot2 book v3 by Hadley Wickham, Danielle Navarro, and Thomas Lin Pedersen
• patchwork website
• Cookbook for R by Winston Chang
• R for Data Science by Hadley Wickham and Garret Grolemund
  • Chapter on Data Visualization
• Data Visualization: A practical introduction by Kieran Healy