Eight visualisation challenges with ggplot2

Hadley Wickham  
@hadleywickham 
Chief Scientist, RStudio
Solving 8 visualisation
challenges with ggplot2
November 2016

View Slide

http://ﬁvethirtyeight.com/features/our-47-weirdest-charts-from-2015/

View Slide

https://ﬂowingdata.com/tag/upshot/

View Slide

1Labelling plots
Solved by Bob Rudis
A problem ignored for too long

View Slide

●
●
●
●
●
●
●
●
●
●
●
●
● ●
●
●
●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ●
●
●
●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ●
●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
● ●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ●
20
30
40
2 3 4 5 6 7
displ
hwy
class
●
●
●
●
●
●
●
2seater
compact
midsize
minivan
pickup
subcompact
suv
Two seaters (sports cars) are an exception because of their light weight
Fuel efficiency generally decreases with engine size
Data from fueleconomy.gov

View Slide

ggplot(mpg, aes(displ, hwy)) +
geom_point(aes(color = class)) +
geom_smooth(se = FALSE, method = "loess") +
labs(
title = "Fuel efficiency generally ...",
subtitle = "Two seaters (sports cars) ...",
caption = "Data from fueleconomy.gov"
)
Accessed with the labs() function

View Slide

2
Axes

View Slide

View Slide

Stages of visualisation system popularity
1. Someone used it and complained about a bug
2. Someone used it in an academic paper
3. Someone used it in a newspaper
4.Someone used it to commit academic fraud
5. So many people use it that google has autocompletes
for bad graphics ideas

View Slide

View Slide

Isenberg, Petra, et al. "A study on dual-scale data charts." IEEE Transactions on
Visualization and Computer Graphics 17.12 (2011): 2469-2478.
https://www.lri.fr/~isenberg/publications/papers/Isenberg_2011_ASO.pdf

View Slide

But...

View Slide

●
●
●
●
●
●
●
●
●
●
●
●
● ●
●
●
●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ●
●
●
●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ●
●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
● ●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ●
20
30
40
6
8
10
12
14
16
18
20
2 3 4 5 6 7
displ
mpg
l / 100 km

View Slide

ggplot(mpg, aes(displ, hwy)) +
geom_point() +
scale_y_continuous(
"mpg",
sec.axis = sec_axis(
~ 235 / .,
name = "l / 100 km",
breaks = seq(2, 20, by = 2)
)
)
Only 1-to-1 transformations are allowed
function(x) {
235 / x
}

View Slide

●
●
●
●
●
●
●
●
●
●
●
●
● ●
●
●
●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ●
●
●
●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ●
●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
● ●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ●
20
30
40
6
8
10
12
14
16
18
20
2 3 4 5 6 7
displ
mpg
l / 100 km

View Slide

Labelling
3 
data

View Slide

●
●
●
●
●
●
●
●
●
●
●
●
● ●
●
●
●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ●
●
●
●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ●
●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
● ●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ● corvette
caravan 2wd
altima
forester awd
toyota tacoma 4wd
jetta
new beetle
20
30
40
2 3 4 5 6 7
displ
hwy
class
●
●
●
●
●
●
●
2seater
compact
midsize
minivan
pickup
subcompact
suv
geom_text()

View Slide

●
●
●
●
●
●
●
●
●
●
●
●
● ●
●
●
●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ●
●
●
●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ●
●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
● ●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ● corvette
caravan 2wd
altima
forester awd
toyota tacoma 4wd
jetta
new beetle
20
30
40
2 3 4 5 6 7
displ
hwy
class
●
●
●
●
●
●
●
2seater
compact
midsize
minivan
pickup
subcompact
suv
geom_label()

View Slide

●
●
●
●
●
●
●
●
●
●
●
●
● ●
●
●
●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ●
●
●
●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ●
●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
● ●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ● ●
●
●
●
●
●
●
corvette
caravan 2wd
altima
forester awd
toyota tacoma 4wd
jetta
new beetle
20
30
40
2 3 4 5 6 7
displ
hwy
class
●
●
●
●
●
●
●
2seater
compact
midsize
minivan
pickup
subcompact
suv
https://github.com/slowkow/ggrepel
geom_label_repel()

View Slide

●
●
●
●
●
●
●
●
●
●
●
●
● ●
●
●
●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ●
●
●
●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ●
●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
● ●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ● ●
●
●
●
●
●
●
corvette
caravan 2wd
altima
forester awd
toyota tacoma 4wd
jetta
new beetle
20
30
40
2 3 4 5 6 7
displ
hwy
class
●
●
●
●
●
●
●
2seater
compact
midsize
minivan
pickup
subcompact
suv
dev version

View Slide

View Slide

Two diﬀerence between a factor and a string:
1.Fixed set of possible values
2.Arbitrary order

View Slide

relig %
group_by(relig) %>%
summarise(
tvhours = mean(tvhours, na.rm = TRUE),
n = n()
)
Some data from the general social survey

View Slide

●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
No answer
Don't know
Inter−nondenominational
Native american
Christian
Orthodox−christian
Moslem/islam
Other eastern
Hinduism
Buddhism
Other
None
Jewish
Catholic
Protestant
2 3 4
tvhours
relig

View Slide

●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
Other eastern
Hinduism
Buddhism
Orthodox−christian
Moslem/islam
Jewish
None
No answer
Other
Christian
Inter−nondenominational
Catholic
Protestant
Native american
Don't know
2 3 4
tvhours
fct_reorder(relig, tvhours)

View Slide

by_age %
filter(!is.na(age)) %>%
group_by(age, marital) %>%
count() %>%
mutate(prop = n / sum(n))
You have the same problem with more dimensions

View Slide

0.00
0.25
0.50
0.75
1.00
20 40 60 80
age
prop
marital
No answer
Never married
Separated
Divorced
Widowed
Married

View Slide

0.00
0.25
0.50
0.75
1.00
20 40 60 80
age
prop
marital
Widowed
Married
Divorced
Never married
No answer
Separated

View Slide

5
Missing
values

View Slide

An explicit missing value (NA)
is the presence of an absence;  
an implicit missing value is the
absence of a presence.

View Slide

Demo

View Slide

6
Histograms

View Slide

hist(1:4)

View Slide

df df %>%
ggplot(aes(x)) +
geom_histogram(binwidth = 1)
Equivalent ggplot2 code is a little longer

View Slide

0.00
0.25
0.50
0.75
1.00
1 2 3 4
x
count
(0.5, 1.5] (1.5, 2.5] (2.5, 3.5] (3.5, 4.5]
Thanks to Randall Pruim

View Slide

df %>%
ggplot(aes(x)) +
geom_histogram(
binwidth = 1,
boundary = 0
)
df %>%
ggplot(aes(x)) +
geom_histogram(
binwidth = 1,
boundary = 0,
closed = "left"
)

View Slide

0.0
0.5
1.0
1.5
2.0
1 2 3 4
x
count
[1, 2] (2, 3] (3, 4]

View Slide

0.0
0.5
1.0
1.5
2.0
1 2 3 4
x
count
(0.5, 1.5]
[1, 2) [2, 3) [3, 4]

View Slide

0.0
0.5
1.0
1.5
2.0
1 2 3 4
x
count
(0.5, 1.5]
[0.99999, 1.99999) [1.99999, 2.99999) [2.99999, 4.00001]

View Slide

7
Bar
charts

View Slide

0
20
40
60
2seater compact midsize minivan pickup subcompact suv
class
count
ggplot(mpg, aes(class)) +
geom_bar(colour = "white")

View Slide

0
20
40
60
class
count
ggplot(mpg, aes(class, group = id)) +
geom_bar(col = "white")

View Slide

0
20
40
60
class
count
drv
4
f
r
ggplot(mpg, aes(class, group = id, fill = drv)) +

View Slide

0
20
40
60
class
count
drv
4
f
r
ggplot(mpg, aes(class, fill = drv)) +

View Slide

class_mpg %
group_by(class) %>%
summarise(
mean = mean(hwy),
se = 1.96 * sd(hwy) / sqrt(n())
)
Another type of bar chart displays summaries

View Slide

0
10
20
class
mean
ggplot(class_mpg, aes(class, mean)) +
geom_bar(stat = "identity")

View Slide

0
10
20
class
mean
ggplot(class_mpg, aes(class, mean)) +
geom_col() # Thanks to Bob Rudis

View Slide

●
●
●
●
●
●
●
20
24
28
class
mean

View Slide

●
●
●
●
●
●
●
15
20
25
30
class
mean

View Slide

8
ggplot2
extension
9 10
11

View Slide

2.1.0 introduced a formal extension mechanism
https://www.ggplot2-exts.org, by Daniel Emaasit

View Slide

ggraph, by Thomas Lin Pedersen
https://github.com/thomasp85/ggraph

View Slide

ggseas by Peter Ellis
https://github.com/ellisp/ggseas
Uses X13-SEATS-ARIMA  
in seasonal package

View Slide

gganimate by David Robinson
https://github.com/dgrtwo/gganimate

View Slide

Conclusion

View Slide

1Labelling plots
Solved by Bob Rudis
A problem ignored for too long

View Slide

2
Axes

View Slide

Labelling
3 
data

View Slide

View Slide

5
Missing
values

View Slide

6
Histograms

View Slide

7
Bar
charts

View Slide

8
ggplot2
extension
9 10
11

View Slide

Many of the features I
discussed here have
been added in recent
versions of ggplot2.  
 
See the release notes
for more detail.

View Slide

http://ggplot2.tidyverse.org

View Slide

Eight visualisation challenges with ggplot2

Eight visualisation challenges with ggplot2

Hadley Wickham

More Decks by Hadley Wickham

Other Decks in Science

Featured

Transcript