CODEWITHSUNDEEP
rggplot2geom-bar
MeganCole
MeganCole

Reputation: 49

order y axis by count of one particular value in column with geom_bar

wondering how I can order the clusters on y-axis by decreasing count of kiwi?

df = data.frame()
df = data.frame(matrix(df, nrow=200, ncol=2))
colnames(df) <- c("cluster", "name")
df$cluster <- sample(20, size = nrow(df), replace = TRUE)
df$fruit <- sample(c("banana", "apple", "orange", "kiwi", "plum"), size = nrow(df), replace = TRUE)

p = ggplot(df, aes(x = as.factor(cluster), fill = as.factor(fruit)))+
  geom_bar(stat = 'count') + 
  theme_classic()+
  coord_flip() +
  theme(axis.text.y = element_text(size = 20),
        axis.title.x = element_text(size = 20),
        axis.title.y = element_text(size = 20),
        axis.text=element_text(size=20)) +
  theme(legend.text = element_text(size = 20)) +
  xlab("Cluster")+
  ylab("Fruit count") +
  labs( fill = "")
p

enter image description here

Upvotes: 2

Views: 352

Answers (4)

Robert Hacken
Robert Hacken

Reputation: 4725

No need to modify the data, just use x = reorder(cluster, fruit=='kiwi', sum) in aes() (instead of as.factor(cluster)).

Barplot sorted according to kiwi

ggplot(df, aes(x = reorder(cluster, fruit=='kiwi', sum), 
               fill = as.factor(fruit))) +
  geom_bar(stat = 'count') + 
  theme_classic() +
  coord_flip() +
  theme(axis.text.y = element_text(size = 20),
        axis.title.x = element_text(size = 20),
        axis.title.y = element_text(size = 20),
        axis.text=element_text(size=20)) +
  theme(legend.text = element_text(size = 20)) +
  xlab('Cluster') +
  ylab('Fruit count') +
  labs(fill = '')

Upvotes: 3

Allan Cameron
Allan Cameron

Reputation: 173928

I would probably do this as a data manipulation before plotting. Note I have moved kiwi to the first position in the stacking order so we can see the bars getting smaller as we move down the y axis.

library(tidyverse)

df %>%
  mutate(cluster = factor(cluster, 
                          names(sort(table(fruit == 'kiwi', cluster)[2,]))),
         fruit = factor(fruit, c('kiwi', 'apple', 'banana', 
                                 'orange', 'plum'))) %>%
  ggplot(aes(x = cluster, fill = fruit))+
  geom_bar(position = position_stack(reverse = TRUE)) + 
  theme_classic()+
  coord_flip() +
  theme(axis.text.y = element_text(size = 20),
        axis.title.x = element_text(size = 20),
        axis.title.y = element_text(size = 20),
        axis.text=element_text(size=20)) +
  theme(legend.text = element_text(size = 20)) +
  scale_fill_manual(values = c('olivedrab', 'yellowgreen', 'yellow2', 
                               'orange2', 'plum4')) +
  xlab("Cluster")+
  ylab("Fruit count") +
  labs( fill = "")

enter image description here

Upvotes: 5

zephryl
zephryl

Reputation: 17154

Compute total kiwis by group, then convert cluster to a factor ordered by this grouped kiwi total. Using dplyr and forcats::fct_reorder():

set.seed(13)
library(dplyr)
library(forcats)

df <- df %>% 
  group_by(cluster) %>% 
  mutate(n_kiwi = sum(fruit == "kiwi")) %>% 
  ungroup() %>% 
  mutate(cluster = fct_reorder(factor(cluster), n_kiwi))

p = ggplot(df, aes(x = cluster, fill = fruit))+
  geom_bar(stat = 'count') + 
  theme_classic()+
  coord_flip() +
  theme(axis.text.y = element_text(size = 20),
        axis.title.x = element_text(size = 20),
        axis.title.y = element_text(size = 20),
        axis.text=element_text(size=20)) +
  theme(legend.text = element_text(size = 20)) +
  xlab("Cluster")+
  ylab("Fruit count") +
  labs( fill = "")
p

Upvotes: 0

Antti
Antti

Reputation: 91

Maybe there would be a more efficient way to do this, but one possibility is to count the number of times kiwi occurs for each cluster, then arrange the cluster varaible by that. Note that in this example there can be NAs for the number of kiwis (so we set those instances to 0).

order <- df %>%
 # count how many times kiwi occurs per cluster
 count(fruit, cluster) %>% filter(fruit == 'kiwi')

df <- df %>%
 # join the counts to the original df by cluster
 left_join(order %>% select(cluster, n)) %>% 
 
  # if na make zero (otherwise NAs appear at the top of the plot)
  mutate(n = ifelse(is.na(n), 0, n),
  # arrange the clusters by n
         cluster = fct_reorder(as.factor(cluster), n))

and then then your plotting function should give the desired output.

Upvotes: 0

Related Questions