如何在ggplot中仅分散重叠点且不引入随机噪声？

需求：x轴为分类变量，y轴为连续变量，绘制带均值线、误差棒的散点图时，C、D组的点出现重叠，要求仅对重叠点进行均匀分散，非重叠点保持居中；不使用jitter（随机抖动）、position_dodge（会应用到所有类别），也不使用geom_dotplot（会丢失y值细节），同时保留y值精度。

原始示例代码 #

library(tidyverse)

A <- c(5.1, 5.2, 4.8)
B <- c(1.3, 2.8, 3.2)
C <- c(4.5, 4.5, 4.5)
D <- c(8.9, 7.6, 7.6)

example <- data.frame(A, B, C, D) %>%
              pivot_longer(c(A,B,C, D),
                           names_to = "Type", 
                           values_to = "Value", 
                           cols_vary = "slowest")

ggplot(example, aes(x = Type, y = Value, fill = Type)) +
  stat_summary(fun = "mean", 
               colour = "black", 
               size = 0.3,
               width = 0.4,
               geom = "crossbar") +
  stat_summary(fun.data = mean_sdl, 
               fun.args = list(mult = 1), 
               geom = "errorbar",
               linewidth = 0.8,
               width = 0.3,
               colour = "black") +
  geom_point(size = 3,
             shape = 21,
             colour = "black",
             stroke = 1)

解决方案代码 #

核心思路是先对数据预处理，仅给同一分类下y值重复的点添加x轴方向的固定偏移量，非重复点保持原x位置不变：

library(tidyverse)

A <- c(5.1, 5.2, 4.8)
B <- c(1.3, 2.8, 3.2)
C <- c(4.5, 4.5, 4.5)
D <- c(8.9, 7.6, 7.6)

example <- data.frame(A, B, C, D) %>%
  pivot_longer(c(A,B,C, D),
               names_to = "Type", 
               values_to = "Value", 
               cols_vary = "slowest") %>%
  # 按分类和y值分组，计算每个组内的点数量，生成偏移量
  group_by(Type, Value) %>%
  mutate(
    n = n(),
    # 仅当组内点数>1时，生成对称的偏移量，系数0.1可根据点大小调整
    x_offset = ifelse(n > 1, 
                      seq(from = -(n-1)/2, to = (n-1)/2, length.out = n) * 0.1,
                      0)
  ) %>%
  ungroup() %>%
  # 将分类变量转为数值，方便添加偏移
  mutate(x_num = as.numeric(Type))

ggplot(example, aes(y = Value, fill = Type)) +
  # 均值线和误差棒仍基于原分类位置绘制
  stat_summary(aes(x = Type),
               fun = "mean", 
               colour = "black", 
               size = 0.3,
               width = 0.4,
               geom = "crossbar") +
  stat_summary(aes(x = Type),
               fun.data = mean_sdl, 
               fun.args = list(mult = 1), 
               geom = "errorbar",
               linewidth = 0.8,
               width = 0.3,
               colour = "black") +
  # 散点使用添加偏移后的数值x轴
  geom_point(aes(x = x_num + x_offset),
             size = 3,
             shape = 21,
             colour = "black",
             stroke = 1) +
  # 将x轴刻度改回原分类名称
  scale_x_continuous(breaks = 1:4, labels = unique(example$Type)) +
  labs(x = "Type")

说明 #

• 偏移量系数0.1可根据点的大小和分类间距调整，避免点超出分类范围
• 均值线和误差棒仍基于原分类位置绘制，确保统计信息的准确性
• 仅对同一分类下y值重复的点应用偏移，非重复点保持居中，完全符合需求

内容的提问来源于stack exchange，提问作者Jo-Maree Courtney