如何为负面表述的Likert问题翻转配色方案？

一、HH包实现方案：拆分绘图拼接 #

HH包的likert()函数默认对所有问题使用统一配色，无法直接给单个问题单独设置配色顺序。要实现需求，最直接的方式是拆分数据分别绘图，再拼接在一起：

1. 加载依赖包

library(HH)
library(gridExtra)

2. 拆分并处理数据
   把原数据拆分为正向问题、负向问题两个子集，同时反转负向问题的Likert选项列顺序（让积极选项对应原配色的粉色）：

# 拆分正向/负向问题数据
pos_data <- section_1[section_1$question == "I know when I have understood a new concept or idea", ]
neg_data <- section_1[section_1$question == "After I finish a test, I can't tell whether I have done well or not until I get the results", ]

# 反转负向问题的选项列顺序（从"Not true at all"到"Very true"）
neg_data <- neg_data[, c("question", "pre_post", "Not true at all", "Not really true", "A little bit true", "Mostly true", "Very true")]

3. 分别绘制图表
   给两个子集用相同的配色，负向问题因列顺序反转，粉色会自动对应积极方向（"Not true at all"）：

# 正向问题图表
fig_pos <- likert(pre_post~.| question,
                  data = pos_data,
                  as.percent = TRUE,
                  ReferenceZero = 3,
                  ylab = "",
                  xlim = c(-100, 100),
                  main = "",
                  col = c("#E6007E","#F380BF","#DDDFE4","#80CFF1","#009FE3"),
                  strip = strip.custom(factor.levels = "I know when I have understood \na new concept or idea"),
                  par.strip.text = list(cex=0.6, lines=5))

# 负向问题图表
fig_neg <- likert(pre_post~.| question,
                  data = neg_data,
                  as.percent = TRUE,
                  ReferenceZero = 3,
                  ylab = "",
                  xlim = c(-100, 100),
                  main = list("Knowledge of Cognition", x=unit(.55, "npc")),
                  col = c("#E6007E","#F380BF","#DDDFE4","#80CFF1","#009FE3"),
                  strip = strip.custom(factor.levels = "After I finish a test, I \ncan't tell whether I have \ndone well or not until I get \nthe results"),
                  par.strip.text = list(cex=0.6, lines=5))

4. 拼接两张图表

grid.arrange(fig_pos, fig_neg, ncol=1)

二、ggplot2实现方案（更灵活） #

ggplot2的自定义能力更强，适合实现这类个性化需求，步骤如下：

1. 加载依赖包

library(tidyverse)

2. 整理数据格式
   把宽格式数据转成长格式，标记问题类型，并对负向问题反转选项的因子顺序：

section_1_long <- section_1 %>%
  # 转成长格式
  pivot_longer(cols = c("Very true", "Mostly true", "A little bit true", "Not really true", "Not true at all"),
               names_to = "response",
               values_to = "count") %>%
  # 计算百分比（按问题和前后测分组）
  group_by(question, pre_post) %>%
  mutate(percent = count / sum(count) * 100) %>%
  ungroup() %>%
  # 标记问题是正向/负向
  mutate(question_type = case_when(
    question == "I know when I have understood a new concept or idea" ~ "positive",
    question == "After I finish a test, I can't tell whether I have done well or not until I get the results" ~ "negative"
  )) %>%
  # 对负向问题反转选项的因子顺序，让积极选项对应粉色
  mutate(response = ifelse(question_type == "negative",
                           factor(response, levels = rev(c("Very true", "Mostly true", "A little bit true", "Not really true", "Not true at all"))),
                           factor(response, levels = c("Very true", "Mostly true", "A little bit true", "Not really true", "Not true at all"))))

3. 绘制自定义Likert图表

ggplot(section_1_long, aes(x = pre_post, y = percent, fill = response)) +
  geom_col(position = "fill", width = 0.7) +
  coord_flip() + # 翻转坐标轴，让问题显示在y轴
  facet_wrap(~question, ncol = 1, scales = "free_y") + # 按问题分面
  scale_fill_manual(values = c("#E6007E","#F380BF","#DDDFE4","#80CFF1","#009FE3")) + # 统一配色
  labs(title = "Knowledge of Cognition",
       y = "Percentage",
       x = "Pre/Post",
       fill = "Response") +
  theme_minimal() +
  theme(strip.text = element_text(size = 8, lineheight = 1.2),
        plot.title = element_text(hjust = 0.5))

这样处理后，粉色（#E6007E）在正向问题对应Very true，在负向问题对应Not true at all，统一代表积极方向，符合视觉直观性需求。

内容的提问来源于stack exchange，提问作者Sophie Elizabeth