I have tried the method from similar problems on this site but still have not been able to figure this out.

I'm using the titanic dataset from kaggle. The end result I am looking for is that within each Pclass factor, I want age.class (the bars) to be sorted from low to high by n. My attempt is below.

library(tidyverse)
library(titanic)


df <- titanic::titanic_train

head(df)

# Start -------------------------------------------------------------------
df = df %>% 
  mutate(has.cabin = if_else(Cabin == '', 0, 1) %>% as.factor(),
         Pclass = Pclass %>% as.factor(),
         age.class = case_when(
           Age < 5  ~ 'baby',
           Age >5 & Age < 12 ~ 'Child',
           Age > 12 & Age < 18 ~ 'Teen', 
           Age > 18 & Age < 25 ~ 'Young Adult',
           Age > 25 & Age <35 ~ 'Mid Adult', 
           Age > 35 & Age < 60 ~  'Adult', 
           Age > 60 ~ 'Elderly',
           TRUE ~ 'Undefined'
                              )
         )

plot.data = df %>% count(has.cabin, Pclass, age.class) 
  

lvls <- unique(plot.data$Pclass[order(plot.data$age.class,-plot.data$n)])
plot.data$age.classv2 = factor(plot.data$age.class, levels=lvls)

plot.data %>% 
  ggplot(., aes(x = Pclass, y = n, fill = age.class))   
  geom_col(position = 'dodge')   
  facet_grid(~ has.cabin)

CodePudding user response：

I don't know of an easy way to order factors within a factor within facets. There are the functions

• forcats::fct_reorder to reorder factors
• tidytext::reorder_within to reorder one factor within a facet

I've used the second one and faceted by Pclass and made 2 plots, one for has.cabin == 0 and one for has.cabin == 1 and afterwards stitched them together.

One need a separate variable for the fill argument because internally, reorder_within generates several variables with the facet name appended. If you don't use the extra variable, then you see these names, see the comments in Julia Silge's blog.

library(titanic)
library(dplyr)
#> 
#> Attaching package: 'dplyr'
#> The following objects are masked from 'package:stats':
#> 
#>     filter, lag
#> The following objects are masked from 'package:base':
#> 
#>     intersect, setdiff, setequal, union
library(tidytext)
library(ggplot2)
library(patchwork)

df <- titanic::titanic_train

df <- df %>% 
  mutate(has.cabin = if_else(Cabin == '', 0, 1) %>% as.factor(),
         Pclass = as.factor(Pclass),
         age.class = case_when(
           Age < 5  ~ 'baby',
           Age >5 & Age < 12 ~ 'Child',
           Age > 12 & Age < 18 ~ 'Teen', 
           Age > 18 & Age < 25 ~ 'Young Adult',
           Age > 25 & Age <35 ~ 'Mid Adult', 
           Age > 35 & Age < 60 ~  'Adult', 
           Age > 60 ~ 'Elderly',
           TRUE ~ 'Undefined'
         )
  )

p1 <- df %>%
  count(has.cabin, Pclass, age.class) %>% 
  filter(has.cabin == "0") %>% 
  mutate(age.class.plot = reorder_within(age.class, n, Pclass),
         Pclass = paste0("Plcass ", Pclass)) %>% 
  ggplot(aes(x = age.class.plot, y = n, fill = age.class))   
  geom_col(position = 'dodge')   
  scale_x_reordered()  
  facet_grid(~ Pclass, scales = "free_x")  
  theme(
    axis.title.x = element_blank(),
    axis.ticks.x = element_blank(),
    axis.text.x = element_blank()
  )  
  labs(title = "has.cabin 0")  
  coord_cartesian(ylim = c(0, 180))

p2 <- df %>%
  count(has.cabin, Pclass, age.class) %>% 
  filter(has.cabin == "1") %>% 
  mutate(age.class.plot = reorder_within(age.class, n, Pclass),
         Pclass = paste0("Plcass ", Pclass)) %>% 
  ggplot(aes(x = age.class.plot, y = n, fill = age.class))   
  geom_col(position = 'dodge')   
  scale_x_reordered()  
  facet_grid(~ Pclass, scales = "free_x")  
  theme(
    axis.title = element_blank(),
    axis.ticks = element_blank(),
    axis.text = element_blank()
  )  
  labs(title = "has.cabin 1")  
  coord_cartesian(ylim = c(0, 180))

p1   p2   plot_layout(guides = "collect")

^{Created on 2022-06-30 by the reprex package (v1.0.0)}

I think in order to reorder a factor within a factor within a facet one would need to adapt reorder_within.