Show percent % instead of counts in charts of categorical variables

Question

I'm plotting a categorical variable and instead of showing the counts for each category value.

I'm looking for a way to get ggplot to display the percentage of values in that category. Of course, it is possible to create another variable with the calculated percentage and plot that one, but I have to do it several dozens of times and I hope to achieve that in one command.

I was experimenting with something like

qplot(mydataf) +
  stat_bin(aes(n = nrow(mydataf), y = ..count../n)) +
  scale_y_continuous(formatter = "percent")

but I must be using it incorrectly, as I got errors.

To easily reproduce the setup, here's a simplified example:

mydata <- c ("aa", "bb", NULL, "bb", "cc", "aa", "aa", "aa", "ee", NULL, "cc");
mydataf <- factor(mydata);
qplot (mydataf); #this shows the count, I'm looking to see % displayed.

In the real case, I'll probably use ggplot instead of qplot, but the right way to use stat_bin still eludes me.

I've also tried these four approaches:

ggplot(mydataf, aes(y = (..count..)/sum(..count..))) + 
  scale_y_continuous(formatter = 'percent');

ggplot(mydataf, aes(y = (..count..)/sum(..count..))) + 
  scale_y_continuous(formatter = 'percent') + geom_bar();

ggplot(mydataf, aes(x = levels(mydataf), y = (..count..)/sum(..count..))) + 
  scale_y_continuous(formatter = 'percent');

ggplot(mydataf, aes(x = levels(mydataf), y = (..count..)/sum(..count..))) + 
  scale_y_continuous(formatter = 'percent') + geom_bar();

but all 4 give:

Error: ggplot2 doesn't know how to deal with data of class factor

The same error appears for the simple case of

ggplot (data=mydataf, aes(levels(mydataf))) +
  geom_bar()

so it's clearly something about how ggplot interacts with a single vector. I'm scratching my head, googling for that error gives a single result.

adding to hadley's comment, converting your data into a data frame using mydataf = data.frame(mydataf), and renaming it as names(mydataf) = foo will do the trick — Ramnath, Sep 13 '10 at 03:44

score 255 · Accepted Answer · edited Jul 16 '18 at 07:49

255

Since this was answered there have been some meaningful changes to the ggplot syntax. Summing up the discussion in the comments above:

 require(ggplot2)
 require(scales)

 p <- ggplot(mydataf, aes(x = foo)) +  
        geom_bar(aes(y = (..count..)/sum(..count..))) + 
        ## version 3.0.0
        scale_y_continuous(labels=percent)

Here's a reproducible example using mtcars:

 ggplot(mtcars, aes(x = factor(hp))) +  
        geom_bar(aes(y = (..count..)/sum(..count..))) + 
        scale_y_continuous(labels = percent) ## version 3.0.0

This question is currently the #1 hit on google for 'ggplot count vs percentage histogram' so hopefully this helps distill all the information currently housed in comments on the accepted answer.

Remark: If hp is not set as a factor, ggplot returns:

edited Jul 16 '18 at 07:49

Tung

26,371
7
91
115

answered Jan 24 '13 at 22:27

Andrew

9,090
8
46
59

14

Thanks for this answer. Any idea on how to do it class-wise ? – WAF Feb 25 '15 at 15:07
4

As .@WAF suggests, this answer does not work with faceted data. See @Erwan's comment in http://stackoverflow.com/questions/22181132/normalizing-y-axis-in-histograms-in-r-ggplot-to-proportion-by-group?lq=1 – LeeZamparo Nov 11 '15 at 20:49
4

You might need to prefix `percent` with the package it's from to get the above to work (I did). `ggplot(mtcars, aes(x = factor(hp))) + geom_bar(aes(y = (..count..)/sum(..count..))) + scale_y_continuous(labels = scales::percent)` – mammykins May 22 '19 at 16:22
4

To get around use of facets use `geom_bar(aes(y = (..count..)/tapply(..count..,..PANEL..,sum)[..PANEL..]))` instead. Each facet should sum to 100%. – JWilliman Aug 14 '19 at 01:07
Wasn't variables with ".." around them replaced with the stat()-command? https://ggplot2.tidyverse.org/reference/stat.html – Magnus Nov 14 '19 at 14:18
Can you use stat() to to something similar? – Magnus Nov 14 '19 at 14:18
@Magnus, see my new answer, using the newer `after_stat()` function. – stragu Feb 08 '21 at 04:24
@mammykins Is there any simple solution for faceted data then? The link was not useful – Julien Sep 21 '22 at 08:31
@stragu Link of your answer? – Julien Sep 21 '22 at 08:34
@Julien: https://stackoverflow.com/a/66095887/1494531 – stragu Sep 22 '22 at 11:45

score 58 · Answer 2 · answered Sep 12 '10 at 16:23

58

this modified code should work

p = ggplot(mydataf, aes(x = foo)) + 
    geom_bar(aes(y = (..count..)/sum(..count..))) + 
    scale_y_continuous(formatter = 'percent')

if your data has NAs and you dont want them to be included in the plot, pass na.omit(mydataf) as the argument to ggplot.

hope this helps.

answered Sep 12 '10 at 16:23

Ramnath

54,439
16
125
152

thanks for the suggestion. I just tried it and get "Error: ggplot2 doesn't know how to deal with data of class factor". By the way, if there's only one vector of values, what would be instead of 'foo'? I don't have column labels. – wishihadabettername Sep 12 '10 at 16:27
see my comment after hadley's comment. – Ramnath Sep 13 '10 at 03:45
minor correction: the second command suggested, should have had quotes around "foo": names(mydataf) = "foo". With this and with data.frame() call, it worked. Thanks! – wishihadabettername Sep 15 '10 at 02:28
37

Note that in ggplot2 version 0.9.0 the `formatter` argument will no longer work. Instead, you'll want something like `labels = percent_format())`. – joran Mar 03 '12 at 00:02
25

And with 0.9.0 you'll need to load the `scales` library before using `percent_format()`, otherwise it won't work. 0.9.0 doesn't automatically load supporting packages anymore. – Andrew Mar 16 '12 at 06:22
Note that `na.omit` will omit all rows with an `NA` in any column, even columns unrelated to your plot. – Maxy-B Jun 05 '13 at 04:10
Can someone point me to an explanation/documentation how the "..count.." stuff works, please? I haven't found it the docs. – JerryWho May 17 '14 at 10:59
1

See `? stat_bin`. It shows what additional columns are added to the data frame by `ggplot2`. All extra columns are of the form `..variable..`. – Ramnath May 17 '14 at 13:42
1

Does it make sense to replace `aes(y = (..count..)/sum(..count..))` with simply `aes(y = ..density..)`? Visually it give very similar (but still different) picture – Alexander Kosenkov Jun 11 '14 at 22:01
7

In ggplot 0.9.3.1.0, you'll want to first load the `scales` library, then use `scale_y_continuous(labels=percent)` as mentioned [in the docs](http://docs.ggplot2.org/current/scale_continuous.html#) – adilapapaya Oct 07 '14 at 22:56
1

Note that if you actually want percents, not fractions, you will need to use something like `geom_bar(aes(y = ((..count..)/sum(..count..))*100))` – CoderGuy123 May 01 '15 at 22:07
@Ramnath, I was trying to use the above code with fill = "another categorical variable", I get the plot but the % on y axis and bar height is not as per the actual percentage for the bar, instead it seems it is taking combination of the fill variable and then showing the height. How can I correct this? Can help by suggesting a solution. I have posted the problem at http://stackoverflow.com/questions/41078480/r-shiny-ggplot-bar-and-line-charts-with-dynamic-variable-selection-and-y-axis-to – user1412 Dec 12 '16 at 11:46
1

For those coming to this after 2018, replace "labels = percent_format()" with "labels = scales::percent" – Gigi Aug 05 '18 at 18:28

score 53 · Answer 3 · answered May 26 '16 at 12:46

53

With ggplot2 version 2.1.0 it is

+ scale_y_continuous(labels = scales::percent)

answered May 26 '16 at 12:46

Fabian Hertwig

1,093
13
27

Does not display the right percentages with facets – Julien Sep 21 '22 at 08:42

score 47 · Answer 4 · answered Mar 21 '17 at 08:22

47

As of March 2017, with ggplot2 2.2.1 I think the best solution is explained in Hadley Wickham's R for data science book:

ggplot(mydataf) + stat_count(mapping = aes(x=foo, y=..prop.., group=1))

stat_count computes two variables: count is used by default, but you can choose to use prop which shows proportions.

answered Mar 21 '17 at 08:22

Olivier Ma

1,269
13
24

3

This is the best answer as of June 2017, works with filling by group and with faceting. – Skumin Jun 29 '17 at 15:52
3

For some reason this doesn't allow me to use the `fill` mapping (no error is thrown, but no fill color is added). – Max Candocia Apr 07 '18 at 03:20
1

@MaxCandocia I had to remove `group = 1` in order to get fill mapping. maybe it helps – tjebo Apr 25 '18 at 18:27
3

If I remove the `group` parameter, though, it does not show the proper percentages, since everything belongs to its own group for each unique x value. – Max Candocia Apr 25 '18 at 19:41

score 27 · Answer 5 · answered Jan 28 '16 at 16:39

If you want percentages on the y-axis and labeled on the bars:

library(ggplot2)
library(scales)
ggplot(mtcars, aes(x = as.factor(am))) +
  geom_bar(aes(y = (..count..)/sum(..count..))) +
  geom_text(aes(y = ((..count..)/sum(..count..)), label = scales::percent((..count..)/sum(..count..))), stat = "count", vjust = -0.25) +
  scale_y_continuous(labels = percent) +
  labs(title = "Manual vs. Automatic Frequency", y = "Percent", x = "Automatic Transmission")

When adding the bar labels, you may wish to omit the y-axis for a cleaner chart, by adding to the end:

  theme(
        axis.text.y=element_blank(), axis.ticks=element_blank(),
        axis.title.y=element_blank()
  )

Rtist · Answer 6 · 2020-12-15T05:25:06.527

9

Note that if your variable is continuous, you will have to use geom_histogram(), as the function will group the variable by "bins".

df <- data.frame(V1 = rnorm(100))

ggplot(df, aes(x = V1)) +  
  geom_histogram(aes(y = 100*(..count..)/sum(..count..))) 

# if you use geom_bar(), with factor(V1), each value of V1 will be treated as a
# different category. In this case this does not make sense, as the variable is 
# really continuous. With the hp variable of the mtcars (see previous answer), it 
# worked well since hp was not really continuous (check unique(mtcars$hp)), and one 
# can want to see each value of this variable, and not to group it in bins.
ggplot(df, aes(x = factor(V1))) +  
  geom_bar(aes(y = (..count..)/sum(..count..)))

edited Dec 15 '20 at 05:25

answered Nov 18 '18 at 13:35

Rtist

3,825
2
31
40

1

Great solution. But you forgot to multiply by 100 to get %, i.e. `geom_histogram(aes(y = 100*(..count..)/sum(..count..)))`. – drT Dec 14 '20 at 10:58
`+scale_y_continuous(labels = scales::percent_format())` to display in nice percent format – Waldi Mar 08 '22 at 14:18

score 8 · Answer 7 · edited Mar 26 '18 at 04:53

Here is a workaround for faceted data. (The accepted answer by @Andrew does not work in this case.) The idea is to calculate the percentage value using dplyr and then to use geom_col to create the plot.

library(ggplot2)
library(scales)
library(magrittr)
library(dplyr)

binwidth <- 30

mtcars.stats <- mtcars %>%
  group_by(cyl) %>%
  mutate(bin = cut(hp, breaks=seq(0,400, binwidth), 
               labels= seq(0+binwidth,400, binwidth)-(binwidth/2)),
         n = n()) %>%
  group_by(cyl, bin) %>%
  summarise(p = n()/n[1]) %>%
  ungroup() %>%
  mutate(bin = as.numeric(as.character(bin)))

ggplot(mtcars.stats, aes(x = bin, y= p)) +  
  geom_col() + 
  scale_y_continuous(labels = percent) +
  facet_grid(cyl~.)

This is the plot:

stragu · Answer 8 · 2021-02-08T23:47:18.330

Since version 3.3 of ggplot2, we have access to the convenient after_stat() function.

We can do something similar to @Andrew's answer, but without using the .. syntax:

# original example data
mydata <- c("aa", "bb", NULL, "bb", "cc", "aa", "aa", "aa", "ee", NULL, "cc")

# display percentages
library(ggplot2)
ggplot(mapping = aes(x = mydata,
                     y = after_stat(count/sum(count)))) +
  geom_bar() +
  scale_y_continuous(labels = scales::percent)

You can find all the "computed variables" available to use in the documentation of the geom_ and stat_ functions. For example, for geom_bar(), you can access the count and prop variables. (See the documentation for computed variables.)

One comment about your NULL values: they are ignored when you create the vector (i.e. you end up with a vector of length 9, not 11). If you really want to keep track of missing data, you will have to use NA instead (ggplot2 will put NAs at the right end of the plot):

# use NA instead of NULL
mydata <- c("aa", "bb", NA, "bb", "cc", "aa", "aa", "aa", "ee", NA, "cc")
length(mydata)
#> [1] 11

# display percentages
library(ggplot2)
ggplot(mapping = aes(x = mydata,
                     y = after_stat(count/sum(count)))) +
  geom_bar() +
  scale_y_continuous(labels = scales::percent)

^{Created on 2021-02-09 by the reprex package (v1.0.0)}

(Note that using chr or fct data will not make a difference for your example.)

score 6 · Answer 9 · answered Jun 11 '14 at 21:21

If you want percentage labels but actual Ns on the y axis, try this:

    library(scales)
perbar=function(xx){
      q=ggplot(data=data.frame(xx),aes(x=xx))+
      geom_bar(aes(y = (..count..)),fill="orange")
       q=q+    geom_text(aes(y = (..count..),label = scales::percent((..count..)/sum(..count..))), stat="bin",colour="darkgreen") 
      q
    }
    perbar(mtcars$disp)

Show percent % instead of counts in charts of categorical variables

9 Answers9

Linked

Related