Percentages on the y-axis of a bar graph within each factor (as opposed to total percentages)

Question

I have used the following code:

long_ov %>%
  filter(variable == "overall_mother_protection" | variable == "overall_child_protection" |
           variable == "overall_child_future") %>%
  ggplot(aes(factor(value,levels = 1:5,
                    labels = c('Sangat Setuju', 'Agak Setuju', 'Tidak Setuju atau Tidak Setuju', 
                               'Agak Tidak Setuju', 'Sangat Tidak Setuju')), 
             fill = factor(variable, levels = c('overall_mother_protection', 'overall_child_protection',
                                                'overall_child_future'),
                           labels = c('Perlindungan Ibu', 'Perlindungan Anak', 'Masa Depan Bagi Anak-Anak')))) + 
  theme_fivethirtyeight() + scale_fill_manual(values=c("#3C4C56", "#DB5628", "#B79D79")) +
  geom_bar(aes(y = (..count..)/sum(..count..)), position = "dodge") + 
  scale_x_discrete(labels = function(x) str_wrap(x, width = 15),
                   drop = FALSE, na.translate = FALSE) +
  scale_y_continuous(labels=percent) +
  ggtitle("Program Yasera menyebabkan peningkatan...", subtitle = "di antara 386 ibu") +
  mh_theme

to produce the following graph:

I can provide a bit more context (and a dput function on my data frame) if it would be useful, but I think my issue is quite simple.

As you can see, the y-axis shows percentages, but the total percentage of all 9 bars on the graph adds up to about 100%. Instead, I need the y axis to represent the percentage of each x-axis outcome for each individual fill variable (as specified within the fill argument of the initial ggplot call). In other words, the blue bars should add up to 100%, the orange bars should add up to 100%... you get it.

I am almost certain that I need to tweak something within the aes argument of my geom_bar layer, but I am not sure exactly how to do it.

(also wouldn't mind a little help displaying the corresponding percentages with text above each bar, but I am working on solving that for myself as well).

score 0 · Answer 1 · answered Dec 13 '22 at 04:22

It looks like what you want to do involves working with relative frequencies so the issue needs to be taken care on the data level.

Disclaimer: Since I do not have a dput of your data I am just "winging it" from what I see in your ggplot code

long_ov %>%
    filter(variable == "overall_mother_protection" | 
           variable == "overall_child_protection" |
           variable == "overall_child_future") %>%
    # See here!
    group_by(variable) %>%
    summarise(n = n()) %>%
    mutate(freq = n / sum(n)) %>%
     ggplot(aes(factor(value,levels = 1:5,
                    labels = c('Sangat Setuju', 'Agak Setuju', 'Tidak Setuju atau Tidak Setuju', 
                               'Agak Tidak Setuju', 'Sangat Tidak Setuju')), 
             fill = factor(variable, levels = c('overall_mother_protection', 'overall_child_protection',
                                                'overall_child_future'),
                           labels = c('Perlindungan Ibu', 'Perlindungan Anak', 'Masa Depan Bagi Anak-Anak')))) + 
  theme_fivethirtyeight() + scale_fill_manual(values=c("#3C4C56", "#DB5628", "#B79D79")) +
  geom_bar(aes(y = freq, position = "dodge") + 
  scale_x_discrete(labels = function(x) str_wrap(x, width = 15),
                   drop = FALSE, na.translate = FALSE) +
  scale_y_continuous(labels=percent) +
  ggtitle("Program Yasera menyebabkan peningkatan...", subtitle = "di antara 386 ibu") +
  mh_theme

score 0 · Answer 2 · answered Dec 13 '22 at 12:00

Much credit to @Bensstats for setting me in the right direction, but here is the code that I needed:

long_ov %>%
  filter(variable == "overall_mother_protection" | 
           variable == "overall_child_protection" |
           variable == "overall_child_future") %>%
  # See here!
  group_by(variable, value) %>%
  summarise(n = n()) %>%
  mutate(freq = n / sum(n)) %>%
  ggplot(aes(x = factor(value,levels = 1:5,
                        labels = c('Sangat Setuju', 'Agak Setuju', 'Tidak Setuju atau Tidak Setuju', 
                                   'Agak Tidak Setuju', 'Sangat Tidak Setuju')), 
             y = freq,
             fill = factor(variable, levels = c('overall_mother_protection', 'overall_child_protection',
                                                'overall_child_future'),
                           labels = c('Perlindungan Ibu', 'Perlindungan Anak', 'Masa Depan Bagi Anak-Anak')))) + 
         theme_fivethirtyeight() + scale_fill_manual(values=c("#3C4C56", "#DB5628", "#B79D79")) +
         geom_col(position = "dodge") + 
         scale_x_discrete(labels = function(x) str_wrap(x, width = 15),
                          drop = FALSE, na.translate = FALSE) +
         scale_y_continuous(labels=percent) +
         ggtitle("Program Yasera menyebabkan peningkatan...", subtitle = "di antara 386 ibu") +
         mh_theme

So first, I had to group by variable and value in order to get a freq variable equal to the proportion of each value within each variable. That is very specific to my data.

Second, because I was now using an actual variable in my dataset for the y-axis (rather than a count or a frequency computed within the ggplot function), I had to change from geom_bar() to geom_col() which is made to hand an x and y aesthetic.

Percentages on the y-axis of a bar graph within each factor (as opposed to total percentages)

2 Answers2