Q-Step: Data visualisation in R
Yi Liu <y.liu3[at]exeter.ac.uk>
09 November 2017
The plotting system in base R
Base R plotting is done primarily by the plot function:
plot(x = df$Age, y = df$Fare,
main = "Your plot title", sub = "Your plot subtitle",
type = "p", col = "#9d0006")
plot(x = df$Age[df$Survived == 1], y = df$Fare[df$Survived == 1],
main = "Scatter plot of 'Age ~ Fare'",
sub = "How can one survive the Titanic accident",
xlab = "Age", ylab = "Fare",
type = "p", col = "#458588")
points(x = df$Age[df$Survived == 0], y = df$Fare[df$Survived == 0],
type = "p", col = "#9d0006")
legend(x = "topright",
legend = c("Survived", "Did not survive"),
lty = 2,
col = c("#458588", "#9d0006"))
Other specialised plots
Histograms
hist(df$Age)
Density plots
density_age <- density(na.omit(df$Age))
plot(density_age)
Alternatively, you can chain the procedure using a %>% pipe:
df$Age %>% na.omit() %>%
density() %>% plot()An example
ggplot(data = df,
mapping = aes(x = Age, y = Fare)) +
geom_point(aes(color = Survived))
Alternatively you can write the above code as
df %>% ggplot(aes(Age, Fare)) +
geom_point(aes(color = Survived))Aesthetics and the layering system
What if we want a scatter plot of “Sex ~ Fare”?
df %>% ggplot(aes(x = Sex, y = Fare)) +
geom_point(aes(color = Survived))
We need to change the main geom (geometric object) to geom_jitter when variables on both axes are categorical.
df %>% ggplot(aes(x = Sex, y = Fare)) +
geom_jitter(aes(color = Survived))
Alternatively we can do a boxplot:
df %>% ggplot(aes(x = Sex, y = Fare)) +
geom_boxplot()
Counts and values
df %>% ggplot(aes(x = Pclass)) +
geom_bar(stat = "count")
df %>% ggplot(aes(x = Pclass)) +
geom_bar(aes(fill = Survived),
stat = "count", position = "stack")
When you need to supply your own y in a barchart:
df %>% group_by(Survived, Pclass) %>%
summarise(AverageFare = mean(Fare, na.rm = TRUE)) %>%
ggplot(aes(x = Pclass, y = AverageFare)) +
# `geom_bar(stat = "identity", ...)` is equivalent to `geom_col(..)`
geom_col(aes(fill = Survived), position = "dodge")
Continuous numerical variables
df %>% ggplot(aes(x = Fare)) +
geom_histogram(aes(fill = Survived), position = "stack")
df %>% ggplot(aes(x = Age)) +
geom_density(aes(fill = Survived), alpha = 0.6)
Line charts
For line charts that represent connections, we ususally need to specify a “group” aesthetics.
df %>%
filter(Embarked != "") %>%
mutate(Embarked = Embarked %>% factor(levels = c("S", "C", "Q"))) %>%
group_by(Embarked, Survived) %>%
summarise(AverageFare = mean(Fare, na.rm = TRUE)) %>%
ggplot(aes(x = Embarked, y = AverageFare,
colour = Survived, group = Survived)) +
geom_point() + geom_line()
Facetting: facet_wrap and facet_grid
df %>% ggplot(aes(x = Age, y = Fare)) +
geom_point(aes(color = Survived)) +
facet_wrap(~ Sex)
facet_wrap allows for flexible column layout:
df %>% ggplot(aes(x = Age, y = Fare)) +
geom_point(aes(color = Survived)) +
facet_wrap(Pclass ~ Sex, ncol = 2)
facet_grid is more ideal for facetting with 2 factors:
df %>% ggplot(aes(x = Age, y = Fare)) +
geom_point(aes(color = Survived)) +
facet_grid(Pclass ~ Sex, scales = "free_y")
Figure decorations
Configurations regarding the figure as a whole are provided by the theme() function. Read all the available options here.
Change the position of the figure legend:
df %>% ggplot(aes(x = Age, y = Fare)) +
geom_point(aes(color = Survived)) +
facet_grid(Pclass ~ Sex, scales = "free_y") +
theme(legend.position = "bottom")
Add title and other elements:
df %>% ggplot(aes(x = Age, y = Fare)) +
geom_point(aes(color = Survived)) +
facet_grid(Pclass ~ Sex, scales = "free_y") +
labs(title = "Place your title here",
subtitle = "Place your subtitle here",
x = "Age of passengers",
y = "Trip fare") +
theme_classic()
Exercise:
Let us practice what we learn today and see if you could reproduce one of the following figures.
What can we learn from Titanic data
Exploratory data analysis assisted by visualisation is only the first step in your analysis.
