如何在 R 中的 ggplot 美学中使用先验中位数？

Question

我正在尝试将中位数计算为一个数字，然后将该数字用作 ggplot 美学中的值。

我首先尝试获取中位数作为值：

mean_delay_median <- nycflights13::flights %>% 
  group_by(dest) %>%
  summarise(mean_delay = mean(arr_delay, na.rm = TRUE)) %>% 
  median(mean_delay)

这会产生错误消息：

Error in median.default(., mean_delay) : need numeric data

我该如何解决这个问题？

一旦我得到这个工作，我的第二步是根据高于和低于这个中值“mean_delay_median”的值为地图着色，如下所示：

nycflights13::flights %>% 
  group_by(dest) %>%
  summarise(mean_delay = mean(arr_delay, na.rm = TRUE)) %>% 
  inner_join(nycflights13::airports, c('dest' = 'faa')) %>% 
  ggplot(aes(lon, lat, color=mean_delay>mean_delay_median)) +
  borders("state") +
  geom_point() +
  coord_quickmap() 

一般来说，我寻求在后续代码中使用先前统计数据的指导。

谢谢！

Answer 1

你只是错过了summarise(median_all_delay = median(mean_delay, na.rm = TRUE))

试试这个：

mean_delay_median <- nycflights13::flights %>% 
  group_by(dest) %>%
  summarise(mean_delay = mean(arr_delay, na.rm = TRUE)) %>% 
  summarise(median_all_delay = median(mean_delay, na.rm = TRUE)) %>% 
  unlist()

nycflights13::flights %>% 
  group_by(dest) %>%
  summarise(mean_delay = mean(arr_delay, na.rm = TRUE)) %>% 
  inner_join(nycflights13::airports, c('dest' = 'faa')) %>% 
  ggplot(aes(lon, lat, color=mean_delay>mean_delay_median)) +
  borders("state") +
  geom_point() +
  coord_quickmap() 

输出是：

Answer 2

您应该注意，一个目的地 (dest) 缺少每个 arr_delay 观察。

library(tidyverse)
library(nycflights13)

flights %>% 
  group_by(dest) %>% 
  filter(all(is.na(arr_delay))) %>% 
  select(dest, arr_delay)
#> # A tibble: 1 x 2
#> # Groups:   dest [1]
#>   dest  arr_delay
#>   <chr>     <dbl>
#> 1 LGA          NA

这导致NaN，而不是零。

mean(c(NA), na.rm = TRUE)
#> [1] NaN

也就是说，你应该在median函数中再次添加na.rm = TRUE。

flights %>% 
  group_by(dest) %>% 
  summarise(mean_delay = mean(arr_delay, na.rm = TRUE)) %>% 
  mutate(arrival = ifelse(mean_delay > median(mean_delay, na.rm = TRUE), "late", "okay")) %>% # na.rm option to median
  inner_join(airports, by = c("dest" = "faa")) %>% 
  ggplot() +
  aes(lon, lat, colour = arrival) +
  borders("state") +
  geom_point() +
  coord_quickmap()

由于LGA 的平均值没有任何值，它的标签可能变为NA。