R中优化Likert与计数条形图：筛选20个负面反馈最多的类别

要实现只展示负面反馈最多的20个类别，核心思路是先筛选目标子集，再基于子集绘图，而非在绘图环节临时处理。以下是修改后的完整代码及说明：

步骤1：计算并筛选前20个负面反馈最多的类别 #

先从原数据中统计每个类别下"Strongly disagree"和"Disagree"的总数量，排序后提取前20个类别：

library(tidyverse)
library(ggstats)
library(ggpubr)

# 保留原数据生成逻辑
set.seed(42)
var_levels <- paste0("Category_", 1:40)
likert_levels <- c("Strongly disagree", "Disagree", "Neither agree nor disagree", "Agree", "Strongly agree")
df <- tibble(
  var = sample(var_levels, 500, replace = TRUE),
  val = sample(likert_levels, 500, replace = TRUE)
) %>% mutate(across(everything(), as.factor))

# 筛选负面反馈Top20类别
top20_negative <- df %>%
  mutate(is_negative = val %in% c("Strongly disagree", "Disagree")) %>%
  group_by(var) %>%
  summarise(negative_count = sum(is_negative)) %>%
  arrange(desc(negative_count)) %>%
  slice_head(n = 20) %>%
  pull(var)

# 生成仅包含目标类别的数据子集
df_filtered <- df %>% filter(var %in% top20_negative)

步骤2：绘制Likert图（基于筛选后子集） #

将筛选后的数据转换为gglikert所需的宽格式，同时按负面反馈数量排序类别：

# 转换为宽格式数据
df2_filtered <- df_filtered %>%
  group_by(var) %>%
  mutate(row = row_number()) %>%
  pivot_wider(names_from = var, values_from = val) %>%
  select(-row)

# 绘制Likert图，按负面反馈总数降序排列
v1 <- gglikert(df2_filtered) +
  aes(y = reorder(.question, 
                  .data[[".answer"]] %in% c("Strongly disagree", "Disagree"), 
                  FUN = sum, decreasing = TRUE))

步骤3：绘制负面反馈计数条形图 #

基于筛选子集统计负面反馈数量，直接生成条形图：

v2 <- df_filtered %>%
  mutate(is_negative = val %in% c("Strongly disagree", "Disagree")) %>%
  group_by(var) %>%
  summarise(negative_count = sum(is_negative)) %>%
  mutate(var = reorder(var, negative_count, decreasing = TRUE)) %>%
  ggplot(aes(y = var, x = negative_count)) +
  geom_bar(stat = "identity", fill = "lightgrey") +
  labs(x = "负面反馈总数")

步骤4：组合两张图表 #

ggarrange(v1, v2, widths = c(6, 2))

关键改动说明 #

• 提前筛选子集：避免全量数据导致图表拥挤，聚焦Top20负面类别
• 统一排序逻辑：Likert图和条形图按同一规则排序，保证视觉一致性
• 简化统计逻辑：单独统计负面反馈数量，比原代码的ifelse写法更直观

内容的提问来源于stack exchange，提问作者Homer Jay Simpson