如何在ggplot2中依据绘图生成的斜率对分面进行排序？

你遇到的问题核心是：after_stat(slope)无法直接用于分面排序——因为after_stat()是在ggplot统计图层（比如geom_smooth）中计算的统计量，而分面的排序逻辑在数据传入ggplot之前就已确定，二者的执行顺序不匹配，所以没法用这个动态计算的统计量来排序分面。

正确的解决思路是提前按分组计算每个id对应的线性回归斜率，再用预计算的斜率值重排id的因子顺序，具体操作如下：

步骤1：分组计算每个id的斜率 #

先对数据按id分组，为每个组单独拟合线性模型，提取斜率值：

slope_df <- data_filtered %>%
  group_by(id) %>%
  summarise(slope = coef(lm(value ~ as.numeric(start_date)))[2], .groups = "drop")

步骤2：合并斜率并重新排序id的因子水平 #

把计算好的斜率合并回原数据，再用fct_reorder()根据斜率值重排id的因子水平（.desc = TRUE表示从高到低排序）：

data_filtered <- data_filtered %>%
  left_join(slope_df, by = "id") %>%
  mutate(id = fct_reorder(id, slope, .desc = TRUE))

步骤3：绘制按斜率排序的分面图 #

现在用重排后的id做分面，就能得到按斜率从高到低排列的分面：

data_filtered %>% 
  ggplot(aes(x = start_date, y = value)) + 
  geom_line() + 
  geom_smooth(method = "lm", se = FALSE) +
  facet_wrap(~id, scales = "free_y") +
  theme_minimal() +
  theme(strip.text.x = element_text(hjust = 1), 
        strip.background = element_blank(),
        strip.placement = "outside", 
        legend.position = "bottom") +
  theme(panel.spacing = unit(1, "lines"))

完整可运行代码 #

整合所有步骤的完整代码：

library(tidyverse)
set.seed(123)
n <- 100

data_filtered <- data.frame(
  start_date = rep(seq(as.Date("2022-01-01"), by = "days", length.out = n), each = 5),
  value = rnorm(n * 5, mean = 10, sd = 3),
  id = rep(letters[1:5], each = n)
)

data_filtered$slope <- ifelse(data_filtered$id == "a", 0.5, -0.5)
data_filtered$value <- data_filtered$value + data_filtered$slope * as.numeric(data_filtered$start_date) / 100

# 分组计算每个id的斜率
slope_df <- data_filtered %>%
  group_by(id) %>%
  summarise(slope = coef(lm(value ~ as.numeric(start_date)))[2], .groups = "drop")

# 合并斜率并重排id因子
data_filtered <- data_filtered %>%
  left_join(slope_df, by = "id") %>%
  mutate(id = fct_reorder(id, slope, .desc = TRUE))

# 绘制分面图
data_filtered %>% 
  ggplot(aes(x = start_date, y = value)) + 
  geom_line() + 
  geom_smooth(method = "lm", se = FALSE) +
  facet_wrap(~id, scales = "free_y") +
  theme_minimal() +
  theme(strip.text.x = element_text(hjust = 1), 
        strip.background = element_blank(),
        strip.placement = "outside", 
        legend.position = "bottom") +
  theme(panel.spacing = unit(1, "lines"))

你之前尝试的fct_reorder(data_filtered, factor(id), slope, na.rm = TRUE, .desc = TRUE)参数有误：fct_reorder的第一个参数应该是要重排的因子变量（即id），且你没有按分组计算每个id的专属斜率，误用了全局斜率值，所以无法实现正确排序。

内容的提问来源于stack exchange，提问作者Cyrus Mohammadian