Principles of Data Visualisation

The step-by-step process of creating graphs in R

Ronald Geskus and Thinh Ong Phuc

Introduction

What to expect

How to make an informative visual representation of data and/or results

• Part I: process of graph creation
• Part II: principles of graph interpretation

Note

Graph = diagram = chart = figure = picture

But it takes

• knowledge of graphics principles
• effort; trial and error
• time

to make a “good” one

Who created this graph?

Answer: Florence Nightingale

Rose plot to convince UK authorities of importance of hygiene

Example of informative graph

Example of noninformative graph

Hinrichsen: risk factors for hepatitis G infection in haemodialysis patients

• Do we need a graph?

  Are there enough data points to merit a plot?

  • Depends on mode of presentation: conference lecture (time limit), article

• If we use a graph, is this one the most truthful and informative?

Graph versus table

Cohort study on 1187 HIV infected persons in The Gambia

Pattern discovery (fake results)

Points to consider

• Why do we make a graph? (purpose)

  1. We want to explore and inspect our data or results

  2. Present data or results, often with a message, to others

• Where do we show the graph: research paper, conference, website

• Who is the audience. Level of background knowledge

• What do we want to show (data)

  • Graph: visual representation of information (“encoding”)

    mapping from data/results to figure

The Grammar of Graphics

The origin

The author

• Leland Wilkinson, The Grammar of Graphics, Springer 2005
• Author of SYSTAT; 1995: sold to SPSS
• 1994 - 2007: Senior VP, SPSS

What is a grammar

• Natural language grammar:
  • Defines components of a sentence (verb, noun, …) and rules how to combine them
  • Does not advise on the choice between synonymous words or whether a sentence is meaningful; one can write grammatically correct nonsense
• Graphics grammar:
  • Defines components of a statistical graph and rules how to combine them
  • Does not advise on the best choice of scale (linear, logarithmic), color or line type; one can create noninformative graphs

• Grammar helps expressing ideas or information in a clear and coherent way

Main components of a graph (R)

• Mapping from information (data) to aesthetic attributes of geometric objects

• Geometic objects: what you see on the plot (points, lines, bars)

• Aesthetic attributes: “role” of variables in plot: location (often along x-axis and y-axis), colour, size, shape

Scatter plot

• “Mapping routine measles vaccination in low-and middle-income countries.” Nature 589, no. 7842 (2021): 415-419. (IF 2023 = 50.5).

• How do you read this plot?

Scatter plot

What information would you need to map each dot correctly onto this plot?

• Country name
• Position
• Size
• Color

Data

label (country)	x (coverage)	y (inequality)	size (population)	color (region)
Nigeria	0.20	-0.07	220M	Sub-Saharan Africa

Data found here: nature_plot.xlsx

# A tibble: 6 × 5
  country          coverage inequality   pop region                             
  <chr>               <dbl>      <dbl> <dbl> <chr>                              
1 Angola              -0.14      0.1    30.2 Sub-Saharan Africa                 
2 Papua New Guinea    -0.35      0.08   16.5 Southeast Asia, East Asia and Ocea…
3 Pakistan             0.12      0.05   82.5 South Asia                         
4 Chad                 0.14      0.01   27.5 Sub-Saharan Africa                 
5 Ethiopia             0.32      0.005  55   Sub-Saharan Africa                 
6 Kenya                0.07     -0.06   30.2 Sub-Saharan Africa                 

Scatter plot

Let’s draw this by hand

Draw this by hand

• Draw the axes
• Draw the data points
• Adjust dot sizes
• Colour-code the regions
• Add key country labels

ggplot

Create Elegant Data Visualisations Using the Grammar of Graphics.

code.r

library(ggplot2)

Draw the axes

code.r

ggplot(df, aes(x = coverage, y = inequality))

Draw the data points

code.r

ggplot(df, aes(x = coverage, y = inequality)) +
  geom_point()

Adjust dot sizes

code.r

ggplot(df, aes(x = coverage, y = inequality)) +
  geom_point(aes(size = pop))

Colour-code the regions

code.r

ggplot(df, aes(x = coverage, y = inequality)) +
  geom_point(aes(size = pop, color = region))

Add key country labels

code.r

ggplot(df, aes(x = coverage, y = inequality)) +
  geom_point(aes(size = pop, color = region)) +
  geom_text(aes(label = country))

Add key country labels

Main components of a graph (R)

• Mapping from data to aesthetic attributes of geometric objects

• Geometic objects: what you see on the plot (points, lines, bars)

• Aesthetic attributes: “role” of variables in plot: location (often along x-axis and y-axis), colour, size, shape

• Scales: map data values to aesthetic attributes

Scales

• Position on x-axis and on y-axis

  May be (summarized) value after statistical transformation

Statistical transformation

• Value transformation (e.g. logarithm)

• Histogram: statistical transformation “sum” or “percentage/frequency” before geometric object “bar” is used

• Linear regression line: statistical transformation: “least squares fit” before geometric object “line” is used

• Boxplot

Scales

• Position on x-axis and on y-axis

  May be (summarized) value after statistical transformation

• Rest: via colour,

  shape (type of point),

  linetype,

  point size

• Reading values from the graph via guides: tick marks and labels, legend

Refine the plot

code.r

cols <- c(
  "Central Europe, Eastern Europe and Central Asia" = "#fa8495",
  "Latin America and Caribbean" = "#4ca258",
  "North Africa and Middle East" = "#6493bb",
  "South Asia" = "#d7c968",
  "Southeast Asia, East Asia and Oceania" = "#7dd8f3",
  "Sub-Saharan Africa" = "#bc5c91"
)

# 1. Don't try to write and run everything at once
# 2. Add one small piece at a time
# 3. If something goes wrong, it's easier to fix

ggplot(df, aes(x = coverage, y = inequality)) +
  geom_point(aes(size = pop, color = region), alpha = 0.8) +
  geom_text(aes(label = country), hjust = -0.2, vjust = 0.2) +
  geom_vline(xintercept = 0, color = "#999999") +
  geom_hline(yintercept = 0, color = "#999999") +
  scale_x_continuous(breaks = c(-0.25, 0, 0.25, 0.5),
                     limits = c(-0.4, 0.55)) +
  scale_y_continuous(breaks = c(-0.1, 0, 0.1), limits = c(-0.16, 0.1)) +
  scale_size_continuous(
    breaks = c(50, 100, 150, 200),
    labels = c("50 million", "100 million", "150 million", "200 million"),
    range = c(0, 8),
    guide = guide_legend(order = 1)
  ) +
  scale_color_manual(
    values = cols,
    guide = guide_legend(order = 2)
  ) +
  labs(x = "Change in MCV1 coverage (2019-2000)",
       y = "Change in absolute geographical inequality (2019-2000)",
       size = NULL,
       color = NULL) +
  theme_classic() +
  theme(
    legend.position = "top",
    legend.direction = "vertical",
    legend.text = element_text(size = 11),
    legend.key.height = unit(0.5, "cm"),
    axis.text = element_text(size = 11)
  )

Refine the plot

Some frequently used geom_*

Histogram and bar charts

geom_*	description	Parameters
geom_histogram()	Create a histogram where x is a continuous variable divided into bins and y is the count of observations in each bin	Required aesthetic: x Additional parameters: bins (number of bins to divide x into) or binwidth (width of each bin)
geom_col()	Create a bar chart from given x and y	Required aesthetic: x and y
geom_bar()	Similar to geom_col() but only x or y is required, the other axis is computed based on the provided stat parameter	Required aesthetic: x or y Additional parameters: stat method to compute the remaining axis (default to stat = "count" in which case remaining axis is the count for each value of given axis)

Some frequently used geom_*

Example: For the following examples, we will use the simulated_covid.rds dataset

simulated_covid <- readRDS("data/simulated_covid.rds") 

head(simulated_covid)

  id  case_name case_type sex age date_onset date_admission   outcome
1  1 jCQH5RSlVq confirmed   m  22 2023-01-01           <NA> recovered
2  2 AdCD3im7sn  probable   m  21 2023-01-08           <NA> recovered
3  3 iDzmfZhFkV  probable   m  21 2023-01-03           <NA> recovered
4  4 sKipHJsjZ2  probable   m  10 2023-01-10           <NA>      died
5  5 xG7GvAjlBf suspected   m  24 2023-01-05           <NA> recovered
6  7 ZWWcBMLzoH confirmed   m  10 2023-01-04           <NA> recovered
  date_outcome date_first_contact date_last_contact   district     outbreak
1         <NA>               <NA>              <NA>   Tan Binh 1st outbreak
2         <NA>         2022-12-31        2023-01-04    Tan Phu 1st outbreak
3         <NA>         2022-12-29        2023-01-05   Binh Tan 1st outbreak
4   2023-01-27         2023-01-10        2023-01-13    Quan 10 1st outbreak
5         <NA>         2023-01-07        2023-01-07    Quan 12 1st outbreak
6         <NA>         2023-01-06        2023-01-07 Binh Thanh 1st outbreak

Some frequently used geom_*

Example: creating a bar chart for number of covid incidence over time

code.r

ggplot(
    data = simulated_covid, 
    aes(x = date_onset)
  ) +                   
  geom_bar(
    color = "cornflowerblue"    
  )+                       
  labs(
    title = "New covid cases over time",
    x = "Date",
    y = "Number of cases"
  )

Some frequently used geom_*

Scatter plot

geom_*	description	Parameters
geom_point()	Create a scatter plot from given x and y	Required aesthetic: x and y

Some frequently used geom_*

Example: creating a scatter plot for number of covid incidence over time

code.r

library(dplyr)
# we need to compute y 
# (case_count before plotting)
aggregated_cases <- simulated_covid |> 
  group_by(date_onset) |> 
  summarize(
    case_count = n()
  )

ggplot(
    aggregated_cases,
    aes(x = date_onset, y = case_count)
  ) +                   
  geom_point(
    color = "cornflowerblue"    
  )+                       
  labs(
    title = "New covid cases over time",
    x = "Date",
    y = "Number of cases"
  )

Some frequently used geom_*

Line chart

geom_*	description	Parameters
geom_line()	Create a line plot from given x and y	Required aesthetic: x and y
geom_smooth()	Create a regression line plot from x and y	Required aesthetic: x and y method: select regression model ("auto", "lm", "glm", "gam", "loess") formula: adjust the formula for the regression model

Some frequently used geom_*

Example: creating a line chart for number of covid incidence over time

code.r

ggplot(
    aggregated_cases,
    aes(x = date_onset, 
        y = case_count)
  ) +                   
  geom_line(
    color = "cornflowerblue"    
  )+                       
  labs(
    title = "New covid cases over time",
    x = "Date",
    y = "Number of cases"
  )

Some frequently used geom_*

Example: add a regression line for number of covid incidence over time

code.r

ggplot(
    aggregated_cases,
    aes(x = date_onset, y = case_count)
  ) +                   
  geom_point(
    color = "grey"    
  )+  
  geom_smooth(
      color = "cornflowerblue",
      # choose gam model to fit the data
      method = "gam" 
    )+
  labs(
    title = "New covid cases over time",
    x = "Date",
    y = "Number of cases"
  )

Some frequently used geom_*

Box plot

geom_*	description	Parameters
geom_boxplot()	Create a boxplot from a continuous variable. If both x and y are given, one must be a categorical variable (in which case a boxplot is created for each category)	Required aesthetic: x or y

Some frequently used geom_*

Example: create a box plot to compare number of cases per day between 2 outbreaks

code.r

# compute cases per day for each outbreak
cases_per_outbreak <- simulated_covid |> 
  group_by(date_onset, outbreak) |> 
  summarize(
    case_count = n()
  ) 

ggplot(
    cases_per_outbreak,
    aes(x = case_count, y = outbreak)
  ) +                   
  geom_boxplot()+
  labs(
    title = "Compare distribution of daily cases between 2 outbreaks",
    y = "Outbreak",
    x = "Number of daily cases"
  )

Where to find more geom_*?

For a complete list of geom_*, refer to ggplot2 documentation

Alternatively, you can also look for the example code for your desired chart type at R Graph Gallery

Save plot

Function ggsave() is used for saving plots with the following key parameters

• plot select plot to save (default to the last generated plot)

• filename name of the saved file

• path path to the directory where the plot is stored

• width, height plot size in units expressed by the units argument

• dpi plot resolution

Save plot

Example: save the last generated plot

ggsave(
  path = "plots", # save plot under plots directory
  filename = "new_plot.png",
  width = 800, height = 600, # adjust plot size in pixel
  units = "px",
  dpi = 200 # set resolution
)

Export: two types of formats

• Vector format (pdf, eps, wmf, emf, svg)
  • digital image consisting of independent geometric objects (segments, polygons, curves, etc.)
  • can be enlarged without losing resolution
• Raster format (png, jpeg, tiff, bmp).
  • rectangular grid of pixels, possibly with color
  • resolution impaired if image is enlarged

Main components of a graph (R)

• Mapping from data to aesthetic attributes of geometric objects

• Geometic objects: what you see on the plot (points, lines, bars)

• Aesthetic attributes: “role” of variables in plot: location (often along x-axis and y-axis), colour, size, shape

• Scales: map data values to aesthetic attributes

• Coordinate system: cartesian, polar, map projections.

  Networks, trees have no coordinate system

• Not about:
  • graph type (\(\approx\) language, words)
  • font size, background colour … Some of these are specified in a theme

Polar coordinate system

Pie chart WHO European Health Report 2009

Geometric object? Aesthetic attributes? Transformation? Coordinate system?

Advance ggplot2

Tuyen Huynh

Overview

• What does mapping = aes() do?
• Inheritance of data and aesthetics
• The annotate() layer
• Splitting up with facet_wrap()
• Extending ggplot2
• Interactive plots with plotly

What does mapping = aes() do?

• ggplot() arguments and default values:

ggplot(data = NULL, mapping = aes(), ...)

• geom_*() arguments and default values:

geom_bar(mapping = NULL, data = NULL, ...)
geom_boxplot(mapping = NULL, data = NULL, ...)
geom_point(mapping = NULL, data = NULL, ...)

What does mapping = aes() do?

Recall this:

df <- read_excel("data/nature_plot.xlsx")
head(df)

# A tibble: 6 × 5
  country          coverage inequality   pop region                             
  <chr>               <dbl>      <dbl> <dbl> <chr>                              
1 Angola              -0.14      0.1    30.2 Sub-Saharan Africa                 
2 Papua New Guinea    -0.35      0.08   16.5 Southeast Asia, East Asia and Ocea…
3 Pakistan             0.12      0.05   82.5 South Asia                         
4 Chad                 0.14      0.01   27.5 Sub-Saharan Africa                 
5 Ethiopia             0.32      0.005  55   Sub-Saharan Africa                 
6 Kenya                0.07     -0.06   30.2 Sub-Saharan Africa                 

label (country)	x (coverage)	y (inequality)	size (population)	color (region)
Nigeria	0.20	-0.07	220M	Sub-Saharan Africa

What does mapping = aes() do?

Idea: we are mapping the variables in the data to the aesthetics of a geometry

aes(x = coverage, y = inequality)
#> Aesthetic mapping: 
#> * `x` -> `coverage`
#> * `y` -> `inequality`

aes(x = coverage^2, y = inequality/2)
#> Aesthetic mapping: 
#> * `x` -> `coverage^2`
#> * `y` -> `inequality/2`

What does mapping = aes() do?

How do we know which aesthetics does each geometry support? Read the documentation!

Simply:

• Read the documentation of the geometry of interest
• Click Aesthetics on the navigation bar on the right (or scroll down)
• All aesthetics compatible for the geometry are listed

Inheritance of data and aesthetics

Recall this:

• ggplot() arguments and default values:

ggplot(data = NULL, mapping = aes(), ...)

• geom_*() arguments and default values:

geom_bar(mapping = NULL, data = NULL, ...)
geom_boxplot(mapping = NULL, data = NULL, ...)
geom_point(mapping = NULL, data = NULL, ...)

Inheritance of data and aesthetics

The idea is that:

• ggplot() can control the dataset and aesthetics
• geom_*()s can also control the type of geometries, and allow customisation of aesthetics

Inheritance of data and aesthetics

A typical ggplot plot will look something like:

ggplot(plot_df, aes(x = date, y = count)) +
  geom_line() +
  geom_point()

Implicitly, you can think of this as:

ggplot() +
  geom_line(mapping = aes(x = date, y = count), data = plot_df) +
  geom_line(mapping = aes(x = date, y = count), data = plot_df)

Note

• We call this inheritance
• Data and aesthetics are inherited top-down

Inheritance of data and aesthetics

You have seen this in action before!

cols <- c(
  "Central Europe, Eastern Europe and Central Asia" = "#fa8495",
  "Latin America and Caribbean" = "#4ca258",
  "North Africa and Middle East" = "#6493bb",
  "South Asia" = "#d7c968",
  "Southeast Asia, East Asia and Oceania" = "#7dd8f3",
  "Sub-Saharan Africa" = "#bc5c91"
)

ggplot(df, aes(x = coverage, y = inequality)) +
  geom_point(aes(size = pop, color = region, alpha = 0.8)) +
  geom_text(aes(label = country), hjust = -0.2, vjust = 0.2) +
  geom_vline(xintercept = 0, color = "#999999") +
  geom_hline(yintercept = 0, color = "#999999") +
  scale_x_continuous(breaks = c(-0.25, 0, 0.25, 0.5),
                     limits = c(-0.4, 0.55)) +
  scale_y_continuous(breaks = c(-0.1, 0, 0.1), limits = c(-0.16, 0.1)) +
  scale_size_continuous(
    breaks = c(50, 100, 150, 200),
    labels = c("50 million", "100 million", "150 million", "200 million"),
    range = c(0, 8),
    guide = guide_legend(order = 1)
  ) +
  scale_color_manual(
    values = cols,
    guide = guide_legend(order = 2)
  ) +
  labs(x = "Change in MCV1 coverage (2019-2000)",
       y = "Change in absolute geographical inequality (2019-2000)",
       size = NULL,
       color = NULL) +
  theme_classic() +
  theme(
    legend.position = "top",
    legend.direction = "vertical",
    legend.text = element_text(size = 11),
    legend.key.height = unit(0.5, "cm"),
    axis.text = element_text(size = 11)
  )

Inheritance of data and aesthetics

ggplot(df, aes(x = coverage, y = inequality)) +
  geom_point(aes(size = pop, color = region), alpha = 0.8) +
  geom_text(aes(label = country), hjust = -0.2, vjust = 0.2) 

Inside and outside aes()

What happens if we do it like this:

ggplot(df, aes(x = coverage, y = inequality)) +
  geom_point(aes(size = pop, color = region, alpha = 0.8)) +
  geom_text(aes(label = country, hjust = -0.2, vjust = 0.2)) 

Inside and outside aes()

Back to this slide (again):

label (country)	x (coverage)	y (inequality)	size (population)	color (region)
Nigeria	0.20	-0.07	220M	Sub-Saharan Africa

df <- read_excel("data/nature_plot.xlsx")
head(df)

# A tibble: 6 × 5
  country          coverage inequality   pop region                             
  <chr>               <dbl>      <dbl> <dbl> <chr>                              
1 Angola              -0.14      0.1    30.2 Sub-Saharan Africa                 
2 Papua New Guinea    -0.35      0.08   16.5 Southeast Asia, East Asia and Ocea…
3 Pakistan             0.12      0.05   82.5 South Asia                         
4 Chad                 0.14      0.01   27.5 Sub-Saharan Africa                 
5 Ethiopia             0.32      0.005  55   Sub-Saharan Africa                 
6 Kenya                0.07     -0.06   30.2 Sub-Saharan Africa                 

Inside and outside aes()

Important

In short:

• You want the aesthetics to change with the data -> put it inside aes()*
• You want the aesthetics to be fixed regardless of data -> put it outside aes()*
• The aesthetics must be supported by the geometry

• This also means that columns names can only be used inside aes()

Color transparency

In ggplot2, you can add transparency to colors with the alpha aesthetics (available for most geometries)

The annotate() layer

What if we want to annotate specific parts of a figure?

ggplot(aggregated_cases, aes(x = date_onset, y = case_count)) +                   
  geom_point(color = "grey") +
  geom_smooth(color = "cornflowerblue", method = "gam") +
  labs(
    title = "New covid cases over time",
    x = "Date",
    y = "Number of cases"
  )

The annotate() layer

Adding a vertical line with geom_vline() and a label with geom_label()

ggplot(aggregated_cases, aes(x = date_onset, y = case_count)) +
  # main plotting geoms
  geom_point(color = "grey") +
  geom_smooth(color = "cornflowerblue", method = "gam") +
  # figure annotation geoms (2)
  geom_vline(xintercept = as.Date("2023-03-01"), linetype = 2) +
  geom_label(label = "Start of outbreak", x = as.Date("2023-03-01"), y = 30) +
  # labels
  labs(
    title = "New covid cases over time",
    x = "Date",
    y = "Number of cases"
  )

The annotate() layer

Adding a region with geom_rect() and a label with geom_label()

ggplot(aggregated_cases, aes(x = date_onset, y = case_count)) +
  # figure annotation geoms (1)
  geom_rect(
    ymin = -Inf, ymax = Inf,
    xmin = as.Date("2023-04-01"), xmax = as.Date("2023-06-01"), 
    alpha = 0.2, fill = "red"
  ) +
  geom_label(label = "Outbreak peak", x = as.Date("2023-06-25"), y = 30, color = "red") +
  # main plotting geoms
  geom_point(color = "grey") +
  geom_smooth(color = "cornflowerblue", method = "gam") +
  # labels
  labs(
    title = "New covid cases over time",
    x = "Date",
    y = "Number of cases"
  )

The annotate() layer

We already said alpha = 0.2 outside of aes(), why doesn’t it have any transparency at all??

The annotate() layer

• I lied
• Aesthetics are always dependant on the dataset, whether it is outside or inside aes()

The annotate() layer

Example:

• Originally the dataset when used in ggplot() looks like this:

ggplot(aggregated_cases, ...) + ...

# A tibble: 257 × 2
   date_onset case_count
   <date>          <int>
 1 2023-01-01          1
 2 2023-01-03          1
 3 2023-01-04          1
 4 2023-01-05          1
 5 2023-01-06          3
 6 2023-01-07          3
 7 2023-01-08          1
 8 2023-01-09          1
 9 2023-01-10          2
10 2023-01-14          1
# ℹ 247 more rows

• When you assigned a fixed value to an aesthetic outside of aes(), it might look like this:

ggplot(aggregated_cases, ...) + geom_...(alpha = 0.2)

# A tibble: 257 × 3
   date_onset case_count alpha
   <date>          <int> <dbl>
 1 2023-01-01          1   0.2
 2 2023-01-03          1   0.2
 3 2023-01-04          1   0.2
 4 2023-01-05          1   0.2
 5 2023-01-06          3   0.2
 6 2023-01-07          3   0.2
 7 2023-01-08          1   0.2
 8 2023-01-09          1   0.2
 9 2023-01-10          2   0.2
10 2023-01-14          1   0.2
# ℹ 247 more rows

The annotate() layer

Using the annotate() layer, you can separate the annotations from the geometry layer

ggplot(aggregated_cases, aes(x = date_onset, y = case_count)) +
  # figure annotation geoms (1)
  annotate("rect",
    ymin = -Inf, ymax = Inf,
    xmin = as.Date("2023-04-01"), xmax = as.Date("2023-06-01"), 
    alpha = 0.2, fill = "red"
  ) +
  annotate("label", label = "Outbreak peak", x = as.Date("2023-06-25"), y = 30, color = "red") +
  # main plotting geoms
  geom_point(color = "grey") +
  geom_smooth(color = "cornflowerblue", method = "gam") +
  # labels
  labs(
    title = "New covid cases over time",
    x = "Date",
    y = "Number of cases"
  )

The annotate() layer

Comparing:

geom_rect(
  ymin = -Inf, ymax = Inf,
  xmin = as.Date("2023-04-01"), xmax = as.Date("2023-06-01"), 
  alpha = 0.2, fill = "red"
) +
geom_label(label = "Outbreak peak", x = as.Date("2023-06-25"), y = 30, color = "red")

vs.

annotate("rect",
  ymin = -Inf, ymax = Inf,
  xmin = as.Date("2023-04-01"), xmax = as.Date("2023-06-01"), 
  alpha = 0.2, fill = "red"
) +
annotate("label", label = "Outbreak peak", x = as.Date("2023-06-25"), y = 30, color = "red") +

The annotate() layer

Important

In short:

• You can create plot annotations with geom_label(), geom_rect(), geom_abline(), etc. (look at ggplot2 website)
• If it’s not working correctly (looks blurry, result not as expected), use annotate() instead; very easy to switch from geom_*() to annotate()

Splitting up with facet_wrap()

• What if you want to split the plot into different panels for each group?
• Have a look at your data, find a “grouping” variable to split the data

Splitting up with facet_wrap()

• For example: Split our simulated_covid by outbreak

ggplot(data = simulated_covid, aes(x = date_onset)) +                   
  geom_bar(fill = "cornflowerblue") + 
  facet_wrap(~outbreak)

Splitting up with facet_wrap()

• Use facet_wrap() and add in the variable (with a ~ in the front)

ggplot(data = simulated_covid, aes(x = date_onset)) +                   
  geom_bar(fill = "cornflowerblue") + 
  facet_wrap(~outbreak)

Splitting up with facet_wrap()

• You can “free” the scales from their original limits

ggplot(data = simulated_covid, aes(x = date_onset)) +                   
  geom_bar(fill = "cornflowerblue") + 
  facet_wrap(~outbreak, scales = "free_x")

Splitting up with facet_wrap()

• You can “free” the scales from their original limits

ggplot(data = simulated_covid, aes(x = date_onset)) +                   
  geom_bar(fill = "cornflowerblue") + 
  facet_wrap(~outbreak, scales = "free_y")

Splitting up with facet_wrap()

• You can “free” the scales from their original limits

ggplot(data = simulated_covid, aes(x = date_onset)) +                   
  geom_bar(fill = "cornflowerblue") + 
  facet_wrap(~outbreak, scales = "free")

Extending ggplot2

• Add specific functionalities to ggplot2 with extensions!
• How to use:
  • Think of a specific figure you want to plot
  • Try your best to plot it with ggplot2
  • If you think ggplot2 cannot do what you want. Have a look here
  • Read their documentation to see how to use them

Extending ggplot2

Some commonly used extensions:

• patchwork: to patch multiple ggplot2 plots together
• gganimate: to create moving/animating plots
• ggrepel: to repel labels on a plot away from each other
• ggdist: to work with distributions

Interactive plots with plotly

• If you need to create interactive ggplots, you can use the plotly package
• You just need to save your ggplot into an object

p <- ggplot(aggregated_cases, aes(x = date_onset, y = case_count)) +
  geom_point(color = "grey") +
  geom_smooth(color = "cornflowerblue", method = "gam") +
  labs(
    title = "New covid cases over time",
    x = "Date",
    y = "Number of cases"
  )

Interactive plots with plotly

• Then use it in the plotly::ggplotly() function

# install.packages("plotly")
library(plotly)
ggplotly(p)

Interactive plots with plotly

• You can also create 3d plots with plotly

plot_ly(
  data = tibble(x = rnorm(10), y = rnorm(10), z = rnorm(10)),
  x = ~x, 
  y = ~y,
  z = ~z
)