CODEWITHSUNDEEP
rggplot2
qdread
qdread

Reputation: 3953

How to make text appear in the corner of a ggplot with log scales

I usually use infinite values in the position aesthetics of ggplot text objects to make labels appear in the corner of the plot regardless of the data scale. I mainly use this when making multi-panel figures which should have letters in each panel to identify each panel in the figure legend. However, this does not seem to work with log scales if I want the label to appear on the left or bottom, since obviously transforming log(-Inf) returns NaN. Is there an easy fix for this? I could do a long workaround but I was hoping there is something easier. Example is below:

notlogdata <- data.frame(x = 1:3,y = 1:3)
ggplot(notlogdata, aes(x = x,y = y)) +
  geom_point() +
  geom_text(data = data.frame(x = -Inf, y = Inf, l = 'a'), aes(label = l), hjust = -0.5, vjust = 1)

logdata <- data.frame(x = 10^(1:3), y = 10^(1:3))
ggplot(logdata, aes(x = x,y = y)) +
  geom_point() +
  geom_text(data = data.frame(x = -Inf, y = Inf, l = 'a'), aes(label = l), hjust = -0.5, vjust = 1) +
  scale_x_log10() +
  scale_y_log10()

First plot with untransformed axes appears fine: plot with untransformed axes

The second plot does not have a label and returns the warning:

Warning messages:
1: In self$trans$transform(x) : NaNs produced
2: Removed 1 rows containing missing values (geom_text).

Upvotes: 1

Views: 1495

Answers (3)

Gregor Thomas
Gregor Thomas

Reputation: 145785

You can take a log of Inf --- log(Inf) is Inf. The issue is you can't take a log of a negative like -Inf. But log(0) is -Inf, so if you set x = 0 it will work as expected.

ggplot(logdata, aes(x = x,y = y)) +
  geom_point() +
  geom_text(
    data = data.frame(x = 0, y = Inf, l = 'a'),
    aes(label = l), hjust = -0.5, vjust = 1
  ) +
  scale_x_log10() +
  scale_y_log10()

enter image description here

Upvotes: 4

baptiste
baptiste

Reputation: 77106

annotation_custom(gTree(children=gList(textGrob("a", hjust=0,x=0,vjust=1,y=1))))

enter image description here

Upvotes: 3

qdread
qdread

Reputation: 3953

It does not appear to be possible to use negative positioning aesthetics, including -Inf, on a log-transformed axis. Here is a solution using cowplot as suggested by zx8754

library(cowplot)

notlogdata <- data.frame(x = 1:3,y = 1:3)
notlogplot <- ggplot(notlogdata, aes(x = x,y = y)) +
  geom_point() 

logdata <- data.frame(x = 10^(1:3), y = 10^(1:3))
logplot <- ggplot(logdata, aes(x = x,y = y)) +
  geom_point() +
  scale_x_log10() +
  scale_y_log10()

plot_grid(notlogplot, logplot, labels=c('a','b'), hjust=-8)

This outputs the plot below:

cowplot image

Upvotes: 0

Related Questions