Is there a fix for jagged, poor-quality text produced by geom_text()?

Question

While adding annotation text to a plot I noticed that geom_text() produced unsightly, jagged text, while annotate() produced smooth, nice-looking text. Does anyone know why this happens and if there's any way to fix it? I know I could just use annotate() here, but there are probably cases where geom_text() is preferable, and I'd like to find a fix. Also, geom_text() can't be intended to give poor-looking text, so either I'm doing something wrong, or I've run into some sort of subtle side effect.

Here's some fake data and the code to produce the graph, plus an image showing the results.

library(ggplot2)
age = structure(list(age = c(41L, 40L, 43L, 44L, 40L, 42L, 44L, 45L, 
        44L, 41L, 43L, 40L, 43L, 43L, 40L, 42L, 43L, 44L, 43L, 41L)), 
        .Names = "age", row.names = c(NA, -20L), class = "data.frame")
ggplot(age, aes(age)) + 
  geom_histogram() +
  scale_x_continuous(breaks=seq(40,45,1)) +
  stat_bin(binwidth=1, color="black", fill="blue") +
  geom_text(aes(41, 5.2, 
            label=paste("Average = ", round(mean(age),1))), size=12) +
  annotate("text", x=41, y=4.5, 
           label=paste("Average = ", round(mean(age$age),1)), size=12)

enter image description here

Using structure is fine, but don't forget to assign: ````age=structure````. I have edited your example. — Dieter Menne, Jun 08 '12 at 16:47
My fix is that I generate all the important elements in `ggplot`, then `ggsave(file="plot.eps")` and edit that with Adobe Illustrator or some other vector graphics program for publication quality figures. — Maiasaura, Jun 08 '12 at 17:00

Brian Diggs · Accepted Answer · 2013-09-23T14:27:15.463

88

geom_text, despite not using anything directly from the age data.frame, is still using it for its data source. Therefore, it is putting 20 copies of "Average=42.3" on the plot, one for each row. It is that multiple overwriting that makes it look so bad. geom_text is designed to put text on a plot where the information comes from a data.frame (which it is given, either directly or indirectly in the original ggplot call). annotate is designed for simple one-off additions like you have (it creates a geom_text, taking care of the data source).

If you really want to use geom_text(), just reset the data source:

ggplot(age, aes(age)) + 
  scale_x_continuous(breaks=seq(40,45,1)) +
  stat_bin(binwidth=1, color="black", fill="blue") +
  geom_text(aes(41, 5.2, 
            label=paste("Average = ", round(mean(age$age),1))), size=12,
            data = data.frame()) +
  annotate("text", x=41, y=4.5, 
           label=paste("Average = ", round(mean(age$age),1)), size=12)

enter image description here

edited Sep 23 '13 at 14:27

answered Jun 08 '12 at 17:00

Brian Diggs

57,757
13
166
188

1

Good to understand how `geom_text()` works. Thanks for your help Brian. – eipi10 Jun 08 '12 at 17:05
1

+1 Brian, an example of a well written & illustrated answer and useful tip re `annotate()`. – Gavin Simpson Jun 08 '12 at 19:45
1

shouldn't it be `mean(age$age)` for `geom_text` since you reset the data? – colinfang Sep 22 '13 at 12:36
1

@colinfang You are correct. Thank you for the catch. Answer has been updated. – Brian Diggs Sep 23 '13 at 14:27
14

Another easy fix for this is to use the arguement `check_overlap = TRUE` within `geom_text` – Dave Gruenewald Oct 03 '17 at 18:10

score 26 · Answer 2 · edited Jun 20 '20 at 09:12

Try `geom_text(..., check_overlap = TRUE)`*

From the docs ?geom_text, check_overlap says:

If TRUE, text that overlaps previous text in the same layer will not be plotted.

library(ggplot2)
age = structure(list(age = c(41L, 40L, 43L, 44L, 40L, 42L, 44L, 45L, 
                             44L, 41L, 43L, 40L, 43L, 43L, 40L, 42L, 43L, 44L, 43L, 41L)), 
                .Names = "age", row.names = c(NA, -20L), class = "data.frame")

ggplot(age, aes(age)) + 
  geom_histogram() +
  stat_bin(binwidth=1) +
  geom_text(aes(41, 5.2, label=paste("Average = ", round(mean(age),1))), 
            size=12, 
            check_overlap = TRUE)

*_{This is essentially the answer that Dave Gruenewald posted in a comment to Brian's excellent answer. I'm just trying to make that answer more visible!}

Is there a fix for jagged, poor-quality text produced by geom_text()?

2 Answers2

Try `geom_text(..., check_overlap = TRUE)`*

Linked

Is there a fix for jagged, poor-quality text produced by geom_text()?

2 Answers2

Try geom_text(..., check_overlap = TRUE)*

Linked

Try `geom_text(..., check_overlap = TRUE)`*