移动空因子水平，同时保持 ggplot2 中非空水平的顺序

Question

我正在尝试使用按因子水平计算的值制作 ggplot2 列图。我希望创建一个图，其中空因子水平显示在图上，但保留在轴的底部。目前，他们处于领先地位。我已经根据计算值重新排序了我的因子，并希望为它们保留它。我将包括示例数据。

library(tidyverse)
data(mtcars)

mtcars %>%
  mutate(cyl = as.factor(cyl),
    cyl = fct_expand(cyl, c("2", "4", "6", "8"))) %>%
  group_by(cyl) %>%
  summarize(meanMPG = mean(mpg)) %>%
  ungroup() %>%
  mutate(cyl = fct_reorder(cyl, meanMPG)) %>%
  ggplot(aes(x = cyl, y = meanMPG)) +
  geom_col() +
  scale_x_discrete(drop = FALSE) +
  coord_flip() # shows empty level "2" on the top

Answer 1

这出人意料地棘手 - 鉴于您需要做的就是正确排列关卡。我在forcats 中找不到任何直接合适的内容，但我们可以编写自己的重新排序函数。

my_reorder <- function (fac, var) {
  fac <- fct_reorder(fac, {{var}})
  l <- levels(fac)
  nonempty <- levels(factor(fac)) # I got this idea from droplevels()
  empty <- setdiff(l, nonempty)
  fct_relevel(fac, empty, nonempty)
  fct_relevel(fac, empty, nonempty)
}

mtcars %>%
  mutate(cyl = as.factor(cyl),
         cyl = fct_expand(cyl, c("2", "4", "6", "8"))) %>%
  group_by(cyl) %>%
  summarize(meanMPG = mean(mpg)) %>%
  ungroup() %>%
  mutate(cyl = my_reorder(cyl, meanMPG)) %>%
  ggplot(aes(x = cyl, y = meanMPG)) +
  geom_col() +
  scale_x_discrete(drop = FALSE, ) +
  coord_flip() # shows empty level "2" on the top

Answer 2

我们可以在group_by中使用.drop = FALSE

library(dplyr)
library(ggplot2)
library(forcats)
mtcars %>% 
  group_by(cyl = fct_expand(as.factor(cyl), c('2', '4', '6', '8')), 
     .drop = FALSE) %>%
  summarize(meanMPG = mean(mpg), .groups = 'drop') %>% 
  arrange(!is.na(meanMPG), meanMPG) %>% 
  mutate(cyl = factor(cyl, levels = cyl)) %>% 
  ggplot(aes(x = cyl, y = meanMPG)) +
    geom_col() + 
    scale_x_discrete(drop = FALSE) +
    coord_flip()

-输出