ggplot2分面绘制有序变量图时如何同时展示组内与全样本百分比

核心是拆分计算逻辑，避免分组维度冲突：先单独计算四个种族各自的组内百分比，再单独计算所有种族合并的全样本百分比，给全样本数据统一打上all的组别标签，最后把两份数据合并后绘图即可。不需要在原单一种族分组的计算链中强行插入mutate步骤，很容易因为分组层级不匹配出问题。

library(tidyverse)

# 测试数据
set.seed(123)
d <- data.frame(
  race = sample(c("White", "Hispanic", "Black", "Other"), 100, replace = TRUE),
  question1 = sample(0:4, 100, replace = TRUE),
  question2 = sample(0:4, 100, replace = TRUE),
  question3 = sample(0:4, 100, replace = TRUE)
)

# 转换为长格式，统一处理所有题项
d_long <- d %>% 
  pivot_longer(cols = starts_with("question"),
               names_to = "question",
               values_to = "response")

# 计算分种族的组内百分比
pct_race <- d_long %>% 
  group_by(question, race, response) %>% 
  summarise(n = n(), .groups = "drop_last") %>% 
  mutate(pct = n/sum(n)*100) %>% 
  ungroup()

# 计算全样本百分比，组别统一标记为all
pct_all <- d_long %>% 
  group_by(question, response) %>% 
  summarise(n = n(), .groups = "drop_last") %>% 
  mutate(pct = n/sum(n)*100,
         race = "all") %>% 
  ungroup()

# 合并数据，调整组别显示顺序
plot_data <- bind_rows(pct_race, pct_all) %>% 
  mutate(race = factor(race, levels = c("White", "Black", "Hispanic", "Other", "all")))

# 绘图
ggplot(plot_data, aes(x = factor(response), y = pct, fill = race)) +
  geom_col(position = position_dodge(width = 0.8), width = 0.7) +
  # 添加百分比数值标签
  geom_text(aes(label = paste0(round(pct, 1), "%")), 
            position = position_dodge(width = 0.8), 
            vjust = -0.3, size = 3) +
  # 按题项分面
  facet_wrap(~question, nrow = 1) +
  labs(x = "题项响应值", y = "组内百分比（%）", fill = "组别") +
  theme_bw()

• 组别顺序可自由调整：修改mutate(race = factor(...))中的水平顺序，就能把all组放到任意位置，比如放在四个种族最前、最后，或者插在两个种族中间，满足和单种族结果并排展示的需求。
• 可做视觉区分：如果需要把全样本组和单种族组明确区分开，可以给all组单独设置边框、透明度等样式，示例代码如下：

ggplot(plot_data, aes(x = factor(response), y = pct, fill = race)) +
  geom_col(position = position_dodge(width = 0.8), width = 0.7, 
           aes(colour = race, linewidth = race)) +
  # 给all组设置黑色粗边框，其余种族用细白边框
  scale_linewidth_manual(values = c("White" = 0.5, "Black" = 0.5, "Hispanic" = 0.5, "Other" = 0.5, "all" = 1.2)) +
  scale_colour_manual(values = c("White" = "white", "Black" = "white", "Hispanic" = "white", "Other" = "white", "all" = "black")) +
  geom_text(aes(label = paste0(round(pct, 1), "%")), 
            position = position_dodge(width = 0.8), 
            vjust = -0.3, size = 3) +
  facet_wrap(~question, nrow = 1) +
  labs(x = "题项响应值", y = "组内百分比（%）", fill = "组别", colour = "组别", linewidth = "组别") +
  theme_bw()

内容的提问来源于stack exchange，提问作者a_todd12