ggplot2: negative bar chart with dual x axis - r

Question

I want to produce chart like this:
I go with this code:
dat <- data.frame(
name1 = c("A", "B", "C", "D", "E"),
name2 = c("F", "G", "H", "I", "J"),
value = c(-12,10,5,-7,-2)
)
ggplot(dat,aes(x = name1,y = value)) +
geom_bar(stat = "identity",
fill = "#465978",
width = 0.3) +
ylim(-50,50) +
geom_hline(yintercept = 0,
color = "#9e9e9e",
alpha = 0.3,
linetype = "dotted") +
coord_flip() +
theme_bw() +
theme(axis.line = element_blank(),
axis.title = element_blank(),
axis.text.x = element_blank(),
axis.text.y = element_text(colour = "#737373"),
axis.ticks = element_blank(),
panel.grid.major = element_blank(),
panel.grid.minor = element_blank(),
panel.border = element_blank(),
panel.background = element_blank())
and ended up like this: (note: there is a line in the middle, if you can't see, may be due to low resolution)
How to add dat$name2 in the right side (i.e. top x axis)?
If you don't mind, maybe you can help me produce more similar chart?

Answer 1

Offering an alternative to r2evans' solution, you can convert a discrete variable to a continuous one by doing match(x, unique(x)). If you then have continuous variables on both axes, it is easy to add a secondary axis.
Here is how you could do that (with some extra decorations based on your request to make the chart more similar).
library(ggplot2)
dat <- data.frame(
name1 = c("A", "B", "C", "D", "E"),
name2 = c("F", "G", "H", "I", "J"),
value = c(-12,10,5,-7,-2)
)
# Probably easiest to define `cont_name1 = match(name1, unique(name1))` in the data
# instead of having to declare it in the `aes()` every time.
ggplot(dat,aes(x = value,y = match(name1, unique(name1)))) +
geom_tile(width = Inf, height = 0.3, fill = "grey95") +
geom_bar(stat = "identity",
fill = "#465978",
width = 0.3, orientation = "y") +
geom_segment(aes(y = match(name1, unique(name1)) - 0.15,
yend = match(name1, unique(name1)) + 0.3,
xend = value)) +
geom_text(aes(x = ifelse(value < 0, value - 4, value + 4),
y = match(name1, unique(name1)) + 0.2,
label = scales::percent(value, scale = 1, accuracy = 1)),
vjust = 0) +
xlim(-50,50) +
scale_y_continuous(
breaks = match(dat$name1, unique(dat$name1)),
labels = dat$name1,
sec.axis = sec_axis(~ .x, labels = dat$name2, breaks = 1:5)
) +
geom_vline(xintercept = 0,
color = "#9e9e9e",
alpha = 0.3,
linetype = "dotted") +
ggtitle("Title here (0%)") +
theme_bw() +
theme(axis.line = element_blank(),
axis.title = element_blank(),
axis.text.x = element_blank(),
axis.text.y = element_text(colour = "#737373"),
axis.ticks = element_blank(),
plot.title = element_text(hjust = 0.5, colour = "grey60"),
panel.grid.major = element_blank(),
panel.grid.minor = element_blank(),
panel.border = element_blank(),
panel.background = element_blank())
Created on 2021-01-17 by the reprex package (v0.3.0)

Answer 2

Normally (with a continuous scale) I'd recommend scale_y_continuous(sec.axis=...). Unfortunately, scale_discrete does not yet support it (see https://github.com/tidyverse/ggplot2/issues/3171). With that, the two ways to go include geom_text and annotations. I'll offer the first.
To make it consistent on both sides, we'll need to remove the axis labels from the left axis. (I find this annoying, but consistency between axes is important to me. If it is not as much to you, then you can reduce some of the changes.)
ggplot(dat,aes(x = name1,y = value)) +
geom_bar(stat = "identity",
fill = "#465978",
width = 0.3) +
ylim(-50,50) +
geom_hline(yintercept = 0,
color = "#9e9e9e",
alpha = 0.3,
linetype = "dotted") +
geom_text(aes(y = -Inf, label = name1), hjust = 0, colour = "#737373") + # new
geom_text(aes(y = Inf, label = name2), hjust = 1, colour = "#737373") + # new
coord_flip() +
theme_bw() +
theme(axis.line = element_blank(),
axis.title = element_blank(),
axis.text.x = element_blank(),
axis.text.y = element_blank(), # change
axis.ticks = element_blank(),
panel.grid.major = element_blank(),
panel.grid.minor = element_blank(),
panel.border = element_blank(),
panel.background = element_blank())
One problem you will run into with multi-line labels (as in your example) is that of alignment. ggplot2 has a confounded sense of hjust=, where it means both direction of the text box from the point and alignment of the text within the text box. So hjust=0 means "text goes to the right of the point, and the text is left-aligned". There does not seem to be an easy way to have the text box go to the right of the point (first geom_text, on the left-edge) yet have the text right-justified. (I'll be happy if somebody can show an easy way to work around this!)
The workarounds to force that left-edge right-align textbox require knowing a priori the dimensions of the plot so that you can hard-position the text box (not y=-Inf, the left-edge) within the plot boundary and hard-code the limits of the plot. (Know that when you form the plot, none of the functions know what the dimensions of the rendered plot will be, in user-points, centimeters, or similar.)
As an aside, we can add a geom_text and geom_tile for a couple more features.
ggplot(dat,aes(x = name1,y = value)) +
geom_tile(height = 90, width = 0.3, fill = "gray90") +
geom_bar(stat = "identity",
fill = "#465978",
width = 0.3) +
ylim(-50,50) +
geom_hline(yintercept = 0,
color = "#9e9e9e",
alpha = 0.3,
linetype = "dotted") +
geom_text(aes(y = -Inf, label = name1), hjust = 0, colour = "#737373") +
geom_text(aes(y = Inf, label = name2), hjust = 1, colour = "#737373") +
geom_text(aes(label = value), vjust = -1.1) +
coord_flip() +
theme_bw() +
theme(axis.line = element_blank(),
axis.title = element_blank(),
axis.text.x = element_blank(),
axis.text.y = element_blank(),
axis.ticks = element_blank(),
panel.grid.major = element_blank(),
panel.grid.minor = element_blank(),
panel.border = element_blank(),
panel.background = element_blank())
. + > ggplot(dat,aes(x = name1,y = value)) +
geom_tile(aes(y = 0), height = 90, width = 0.3, fill = "gray90") +
geom_bar(stat = "identity",
fill = "#465978",
width = 0.3) +
ylim(-50,50) +
geom_hline(yintercept = 0,
color = "#9e9e9e",
alpha = 0.3,
linetype = "dotted") +
geom_text(aes(y = -Inf, label = name1), hjust = 0, colour = "#737373") +
geom_text(aes(y = Inf, label = name2), hjust = 1, colour = "#737373") +
geom_text(aes(label = value), vjust = -1.1) +
coord_flip() +
theme_bw() +
theme(axis.line = element_blank(),
axis.title = element_blank(),
axis.text.x = element_blank(),
axis.text.y = element_blank(),
axis.ticks = element_blank(),
panel.grid.major = element_blank(),
panel.grid.minor = element_blank(),
panel.border = element_blank(),
panel.background = element_blank())