如何实现ggplot2两个facet分面中boxplot箱线图各自独立重排序

问题原因 #

你原有代码的排序逻辑是基于全数据集的甲醛中位数计算的，ggplot的分类变量因子水平为全局统一值，默认无法为不同分面设置独立的分类轴排序，可通过以下两种方案解决：

方案1：无额外依赖的基础实现方案 #

通过拼接传感器名和季节名生成唯一的x轴变量，手动设置因子水平后隐藏标签后缀实现视觉上的分面独立排序：

library(tidyverse)

# 1. 数据预处理，生成带排序的拼接变量
DATALL_sort <- DATALL %>%
  # 按季节+传感器分组计算甲醛中位数
  group_by(season, sensor) %>%
  summarise(med_formaldehyde = median(Formaldehyde), .groups = "drop_last") %>%
  # 每个季节内按中位数升序排列
  arrange(med_formaldehyde, .by_group = TRUE) %>%
  # 生成传感器+季节的唯一拼接变量，并设置对应因子水平
  mutate(sensor_season = factor(paste0(sensor, "_", season), 
                                levels = paste0(sensor, "_", season))) %>%
  ungroup() %>%
  # 将排序结果关联回原数据集
  right_join(DATALL, by = c("season", "sensor"))

# 2. 绘图
ggplot(DATALL_sort, aes(x = sensor_season, y = Formaldehyde, fill = sensor)) +
  geom_jitter(position = position_jitter(width = 0.3, height = 0.2), 
              aes(colour = sensor), alpha = 0.4) +
  geom_boxplot(outlier.shape = NA) +
  facet_wrap(~season, scales = "free_x", ncol = 1) +
  scale_y_continuous(limits = quantile(DATALL$Formaldehyde, c(0.1, 0.98))) +
  # 去掉x轴标签的季节后缀，仅显示传感器名
  scale_x_discrete(labels = ~stringr::str_remove(., "_.+$")) +
  labs(x = "", y = expression(Formaldehyde~(30~min)~(µg/ m^{3}))) +
  theme(legend.position = "none")

方案2：使用ggh4x包的简化实现方案 #

ggh4x是ggplot的拓展包，支持直接为分面设置独立的分类轴水平，代码更简洁：

library(tidyverse)
library(ggh4x)

ggplot(DATALL, aes(x = sensor, y = Formaldehyde, fill = sensor)) +
  geom_jitter(position = position_jitter(width = 0.3, height = 0.2), 
              aes(colour = sensor), alpha = 0.4) +
  geom_boxplot(outlier.shape = NA) +
  facet_wrap(~season, scales = "free_x", ncol = 1) +
  # 为每个分面设置独立的x轴排序规则
  scale_x_facet(
    facets = vars(season),
    levels = ~reorder(.$sensor, .$Formaldehyde, FUN = median)
  ) +
  scale_y_continuous(limits = quantile(DATALL$Formaldehyde, c(0.1, 0.98))) +
  labs(x = "", y = expression(Formaldehyde~(30~min)~(µg/ m^{3}))) +
  theme(legend.position = "none")

内容的提问来源于stack exchange，提问作者Maria