Faceting

Create a plot with multiple subplots (facets)

Objectives

• Learn the syntax to make subplots real quick in {ggplot2}
• Learn arguments to modify the appearance of the subplots

Introduction

This satellite builds on the core epicurve session, which is a prerequisite. In that session, we learned how to create an epicurve of measles cases across time:

Now, this plot is cool, but in your sitrep you would like to show the data by age group. There are several ways to do that:

• You could, for each age group, filter your data frame and copy and paste the plotting command to create specific plots
• You could learn to use for loops or apply() or map() family functions, which are very useful ways to automatize actions, and involve less copy and pasting
• Or you could trust {ggplot2} to have a solution…

The first option is tedious and it is error prone, and we advise against it; learning the tools of the second option will be a good investment of you time at some point as they are really powerful, but they are way out of the scope of this tutorial because a much simpler option already exist in {ggplot2}.

Setup

We will use the same clean linelist that we used in the past sessions, which you can download here:

Make sure this dataset is saved into the appropriate subdirectory of your R project and create a new script called faceting.R in your R directory. Add an appropriate header and load the following packages: {here}, {rio}, and {tidyverse}.

Finally, add an import section where you use {here} and {rio} to load your data into an object called df_linelist.

Faceting

The function facet_wrap() allows you to replicate a graph based on the categories of a variable. For example, you could make the epicurve graph by sex, or by site. As other layers of a ggplot graph, you add it to your existing graph with a +. It creates a a figure with multiple small graphs, that {ggplot2} calls facets or small multiples.

Get the Data Ready

In the following session, we will explain the code by creating subplots by sub-prefecture, and you will be plotting the epicurve by age group.

If we want to to plot anything by sub-prefecture, the sub_prefecture variable must be present in the aggregated data frame that we use to plot.

Let’s create a new summarized dataset that has the number of patients by day and by sub-prefecture!

df_pref <- df_linelist %>%
  count(date_onset, sub_prefecture,
        name = "patients")

# A tibble: 6 × 3
  date_onset sub_prefecture patients
  <date>     <chr>             <int>
1 2022-08-13 Moissala              1
2 2022-08-17 Moissala              1
3 2022-08-18 Moissala              1
4 2022-08-22 Moissala              1
5 2022-08-30 Moissala              2
6 2022-09-01 Moissala              1

You will draw a plot of the number of admissions by age group, so you need a new data frame summarized by day and age group. Create this data frame, and call it df_age. It should have this format:

# A tibble: 6 × 3
  date_onset age_group        n
  <date>     <fct>        <int>
1 2022-08-13 1 - 4 years      1
2 2022-08-17 5 - 14 years     1
3 2022-08-18 < 6 months       1
4 2022-08-22 6 - 8 months     1
5 2022-08-30 < 6 months       1
6 2022-08-30 6 - 8 months     1

Add the Facet Layer to the Plot

Now, let’s plot this data. Look at the code bellow: it is exactly the same as before but for the last line, which creates the facets:

df_pref %>%
  ggplot(aes(x = date_onset,
             y = patients)) +
  geom_col(fill = "#2E4573") +
  labs(x = "Date of onset",
       y = "Measles cases",
       title = "Measles cases in Mandoul (Chad)") +
  theme_classic(base_size = 15) +
  facet_wrap(vars(sub_prefecture))   # Make the plot by sub-prefecture!

Isn’t that incredible? As you can see, the function facer_wrap() takes as argument a variable name wrapped in the vars() function.

Now is your turn, draw the epicurve by age group (still keeping all the plots improvement: labels, themes etc.)

It should look like this:

Customize Facets

Check out the function help page to learn about the arguments that facet_wrap() accepts. We will cover a couple here.

Number of Rows or Columns

The arguments nrow and ncol allow you to decide how many facets there should be on one row, respectively one column.

For exemple, we could have all plots on two rows, for a wide figure:

df_pref %>%
  ggplot(aes(x = date_onset,
             y = patients)) +
  geom_col(fill = "#2E4573") +
  labs(x = "Date of onset",
       y = "Measles cases",
        title = "Measles cases in Mandoul (Chad)") +
  theme_classic(base_size = 15) +
  
  facet_wrap(vars(sub_prefecture),
             nrow = 2)  

Or force the number of rows to four, which forces a taller figure:

df_pref %>%
  ggplot(aes(x = date_onset,
             y = patients)) +
  geom_col(fill = "#2E4573") +
  labs(x = "Date of onset",
       y = "Measles cases",
       title = "Measles cases in Mandoul (Chad)") +
  theme_classic(base_size = 15) +
  
  facet_wrap(vars(sub_prefecture),
             nrow = 4)  

Using one of the mentioned argument, create a graph with three columns.

Axis Ranges

Did you notice that in the graph we produced, all facets share the same axis in x and y? This is often a desired feature, as playing with axes is one of the best ways to mislead readers.

That being said, if you are more interesting in seeing the shape of the epicurve by category and less by comparing categories to each other, zooming on the available data can be appropriate (alert your reader to the scale variation though!)

The scales argument accepts the following strings:

• "fixed": the default, same limits on x and y for all facets
• "free_x": the x axis may have different limits in different facets
• "free_y": the y axis may have different limits in different facets
• "free": both axis may vary in different facets

Look at this graph:

We kept time window on the x axis fixed but allowed the y axis to vary to better read the number of cases by sub-prefecture.

Your turn! Draw you graph with age group as facets with a free y axis, and a fixed x axis.

Done!

Very well done team! You have learned how to facet plots! This will work not only on bar plots such as epicurves, but also on other types of plots made by {ggplot2}.

Depending on the size of your graph, the date labels on the x-axis may be a bit messed up, the ones in my examples definitely are. Fear not, this can be controlled and is the object of another satellite!