Adding percentage labels in barplots (gglot2)

Question

I have the following dataset with the following variables indicating whether a person used their phone (a dummy variable with 1 = used the phone ("Yes") and 0 ("No") else); their ID and district and sub-district they live in. Note that a same person may have been recorded twice or more under different sub-districts. However, I only want to count such a person once, that is, consider only unique IDs.

district sub_district   id  used_phone
    A   SX  1   Yes
    A   SX  2   Yes
    A   SX  3   No
    A   SX  4   No
    A   SY  4   No
    A   SY  5   Yes
    A   SZ  6   Yes
    A   SX  6   Yes
    A   SZ  7   No
    B   RX  8   No
    B   RV  9   No
    B   RX  9   No
    B   RV  10  Yes
    B   RV  11  Yes
    B   RT  12  Yes
    B   RT  13  Yes
    B   RV  13  Yes
    B   RT  14  No
    B   RX  14  No
  

N.B: used_phone is a factor variable

For the above dataset, I want to plot a distribution of "whether a person used a phone" for which I was using the following code:

  ggplot(df, aes(x=used_phone)) +
  geom_bar(color = "black", fill = "aquamarine4", position = "dodge") +
  labs(x="Used phone", y = "Number of people") +
  ggtitle("Whether person used phone") +
  theme_bw() +
  theme(plot.title = element_text(hjust = 0.5)))
  

This code works fine. However, I want to do two things:

1. Add % labels for each group (yes & no) over the respective bars but y-axis to show the "count"
2. Plot the graph such that it only considers the unique IDs

Looking forward to solving this with your help as I am novice in R.

Thanks, Rachita

Answer 1

Here is one suggestion that could work:

1. Summarize your df based on used_phone and count total number of people who have either used phone and not.
2. Based on the summarized count, you can calculate percent share and with that you can add label cloumn which is just percent share with % sign
3. You can plot using ggplot and using the new summarized df. You can use geom_text() to add percentage labels at the top of bars, use vjust argument in position_stack() to play around with label's position.

df %>% 
distinct(.keep_all = T) %>%
  group_by(used_phone) %>% 
  summarize(n()) %>% 
  setNames(., c('used_phone', 'count')) %>% 
  mutate('share' = count/sum(count),
         'label' = paste0(round(share*100, 2), '%')) -> df

  ggplot(df, aes(y=count, x=used_phone)) +
  geom_bar(stat='identity',
           color = "black", 
           fill = "aquamarine4", 
           position = "dodge") +
  geom_text(aes(label = label),
            position = position_stack(vjust = 1.02),
            size = 3) +
  labs(title = 'Whether person used phone',
       x = 'Used Phone',
       y = 'Number of People') +
  theme_bw()

Answer 2

As the duplicates in id are id's living in different sub_district at the same time and you want to not double count them, I delete the variable sub_district. Then erase all duplicates, count the phones and calculate the percentage. The DF coming from this is shown. ggplot is with geom_col and the percentage on the axis with scales.

I have commented out two lines of code which allows you to facet for district in your ggplot. The diagram coming out of this is attached at the bottom.

library(tidyverse)

df <- read.table(text="district sub_district   id  used_phone
    A   SX  1   Yes
    A   SX  2   Yes
    A   SX  3   No
    A   SX  4   No
    A   SY  4   No
    A   SY  5   Yes
    A   SZ  6   Yes
    A   SX  6   Yes
    A   SZ  7   No
    B   RX  8   No
    B   RV  9   No
    B   RX  9   No
    B   RV  10  Yes
    B   RV  11  Yes
    B   RT  12  Yes
    B   RT  13  Yes
    B   RV  13  Yes
    B   RT  14  No
    B   RX  14  No", header = T)
table(df$used_phone)
#> 
#>  No Yes 
#>   9  10

ddf <- df %>%
  select(-sub_district) %>%        # delete sub_district
  distinct(id, .keep_all = T) %>%  # unique id`s`
  #group_by(district) %>% 
  count(used_phone) %>%            # cout phones
  mutate(pct = n / sum(n))         # calculate percentage

ddf
#> # A tibble: 2 x 3
#>   used_phone     n   pct
#>   <chr>      <int> <dbl>
#> 1 No             6 0.429
#> 2 Yes            8 0.571

ggplot(ddf, aes(used_phone, pct, fill = used_phone)) +
  geom_col(position = 'dodge') + 
  #facet_wrap(~district) +
  scale_fill_manual(values = c("aquamarine4", "aquamarine3")) +
  scale_y_continuous(labels = scales::percent_format())

---

New Addition based on comment:

• wants y-axis in counts
• wants percentage as labels over the bar
• wants as facet for district

ddf <- df %>%
  select(-sub_district) %>%        # delete sub_district
  distinct(id, .keep_all = T) %>%  # unique id`s`
  group_by(district) %>% 
  count(used_phone) %>%            # cout phones
  mutate(pct = n / sum(n),         # calculate percentage
         label = paste0(round(pct*100, 2), '%'))     

ggplot(ddf, aes(used_phone, n, fill = used_phone)) +
  geom_col(position = 'dodge') + 
  facet_wrap(~district) +
  scale_fill_manual(values = c("aquamarine4", "aquamarine3")) +
  geom_text(aes(label = label),
           position = position_stack(vjust = 1.05),
           size = 3) +
  labs(y='count')

---

*new addition* change the basis for percent

ddf <- df %>%
  select(-sub_district) %>%        # delete sub_district
  distinct(id, .keep_all = T) %>%  # unique id`s`
  mutate(ssum = n()) %>% 
  group_by(district) %>% 
  count(used_phone, ssum) %>%            # cout phones
  mutate(pct = n / ssum,         # calculate percentage
         label = paste0(round(pct*100, 2), '%'))

I have introduced a new variable which sums the numbers up before grouping. That gives: