使用stat_dots按第三变量着色导致分位数点图分布拆分问题

问题出在color映射会自动加入分组依据，导致每个变量下的不同相关强度被拆分为独立组，进而被分散排列。只需强制指定分组仅基于variable，同时保留颜色映射即可解决：

library(dplyr)
library(ggplot2)
library(ggdist)

set.seed(123)
data <- data.frame(
  variable = rep(c("Variable1", "Variable2", "Variable3"), each = 300),
  correlation = c(rnorm(300, 0.2, 0.1), rnorm(300, 0.4, 0.1), rnorm(300, 0.6, 0.1))
)

data <- data %>%
  mutate(correlation_strength = case_when(
    abs(correlation) < 0.3 ~ "Small",
    abs(correlation) >= 0.3 & abs(correlation) < 0.5 ~ "Medium",
    abs(correlation) >= 0.5 ~ "Strong"
  ))

# 核心修改：明确指定分组为variable，颜色仅用于映射不参与分组
ggplot(data, aes(x = variable, y = correlation, group = variable)) +
  stat_dots(aes(color = correlation_strength), quantiles = 50) + 
  scale_color_manual(values = c("Small" = "blue", "Medium" = "green", "Strong" = "red")) +
  theme_minimal() +
  labs(title = "Posterior Draws of Correlation Coefficients",
       subtitle = "Dots colored by correlation strength",
       x = "Variable",
       y = "Correlation Coefficient",
       color = "Correlation Strength")

关键说明 #

• 全局aes中指定group=variable，强制stat_dots仅按变量分组计算点的排列位置，避免color变量干扰分组逻辑。
• 将color=correlation_strength移至stat_dots的aes内部，仅实现颜色映射，不参与分组拆分。
• 移除了position=position_dodge(width=0.8)，默认的点排列逻辑会在每个x轴类别下紧凑排列所有点，无需额外 dodge。

内容的提问来源于stack exchange，提问作者bfoste01