在ggplot中实现均值+标准差的双尺寸点可视化需求

我有一个ggplot图表，用来展示三个及以上变量：用点的大小表示第三个变量time（均值类数值）。现在想添加透明的外层点，其尺寸对应标准差std的相对大小，且外层点的尺寸只体现std的相对关系，不受内层点（对应time）的尺寸影响——避免直接添加geom_point()时，因std的最小尺寸限制导致的视觉偏差。预期效果：无论内层均值点大小如何，小标准差对应较小的外层点，大标准差对应较大的外层点。

示例数据 #

df_experiment <- structure(list(n = c(1, 2, 3, 4, 5, 6, 7, 8, 1, 2, 1, 2, 3, 4, 
5, 6, 7, 8, 9, 10, 1, 2, 3, 4), time = c(10080, 10095, 10110, 
10125, 0, 15, 0, 15, 120, 135, 60, 75, 100, 115, 130, 145, 0, 
15, 0, 15, 60, 75, 45, 60), weight = c(15.5, 23.8, 15.7, 23.3, 
13, 22, 13, 21, 16, 24, 19, 25, 21, 27, 21, 27, 22, 29, 23, 29, 
25, 26, 25, 26), height = c(26.61, 25.23, 23.03, 23.64, 13.64, 
11.84, 13.26, 10.77, 14.01, 15, 13, 13.3, 27.5, 25.92, 18.83, 
19.26, 18.93, 21.43, 16.11, 19.5, 13.01, 13.35, 15.12, 12.92), 
    length = c(47.8, 65.8, 36.7, 63.8, 63, 68, 62, 70.6, 59.8, 
    63.7, 69.3, 66.5, 63.6, 74.2, 67.8, 77.9, 68.1, 78.2, 75.6, 
    81, 67.5, 68.7, 67.5, 68.7), sample = structure(c(1L, 1L, 
    2L, 2L, 3L, 3L, 4L, 4L, 1L, 1L, 1L, 1L, 2L, 2L, 3L, 3L, 4L, 
    4L, 5L, 5L, 1L, 1L, 2L, 2L), .Label = c("A", "B", "C", "D", 
    "E"), class = "factor"), height_set = structure(c(3L, 3L, 
    3L, 3L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 2L, 2L, 2L, 2L, 2L, 
    2L, 2L, 2L, 1L, 1L, 1L, 1L), .Label = c("100", "40", "70"
    ), class = "factor"), weight_set = structure(c(1L, 1L, 1L, 
    1L, 1L, 1L, 1L, 1L, 2L, 2L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 3L, 
    3L, 3L, 3L, 3L, 4L, 4L), .Label = c("10", "15", "20", "25"
    ), class = "factor"), std = c(0.812703886069357, 0.782337151234969, 
    0.602131422609091, 0.69466593535617, 0.580281280446798, 0.430830358527601, 
    0.768334821797907, 0.577222188469023, 0.683016336522996, 
    0.277622280176729, 0.566097403643653, 0.845317166764289, 
    0.328095777193084, 0.83994737197645, 0.874583127908409, 0.424604137195274, 
    0.023338885512203, 0.0864457285497338, 0.929015822941437, 
    0.0764572839252651, 0.194908467587084, 0.65600896649994, 
    0.233351909788325, 0.255731456680223)), row.names = c(NA, 
24L), class = "data.frame")

注：df_experiment$std为随机生成的示例数据

当前绘图代码 #

library(ggplot2)
library(tidyquant) # 用于theme_tq()

df_experiment %>%
  ggplot() + 
  geom_point(aes(x = weight, y = height, color = length, size = time), alpha = 0.8) +
  scale_color_continuous(trans='reverse') +
  scale_size_continuous(range = c(3,9)) +
  xlab("weight") + ylab("height") +
  ggtitle("length") +
  theme_tq()

核心思路：将std做归一化处理，让其尺寸范围独立于time的尺寸范围，确保外层点只反映std的相对大小；调整图层顺序，让内层点（对应time）覆盖在外层点之上。

完整代码 #

library(ggplot2)
library(tidyquant)
library(scales) # 用于rescale函数

df_experiment %>%
  # 归一化std到指定范围（比如4-12，可根据需求调整）
  mutate(std_scaled = rescale(std, to = c(4, 12))) %>%
  ggplot() +
  # 先画外层透明点：尺寸用归一化后的std，低透明度，颜色可选灰色避免干扰
  geom_point(aes(x = weight, y = height, size = std_scaled), 
             alpha = 0.3, color = "gray50") +
  # 外层点的尺寸不需要图例，隐藏它
  scale_size_identity(guide = "none") +
  # 再画内层点：保持原有映射
  geom_point(aes(x = weight, y = height, color = length, size = time), alpha = 0.8) +
  scale_color_continuous(trans='reverse') +
  scale_size_continuous(range = c(3,9)) +
  xlab("weight") + ylab("height") +
  ggtitle("length") +
  theme_tq()

关键说明 #

• 归一化std：用scales::rescale()把原始std缩放到固定范围（示例中是4-12），这样不管std的原始数值大小，其相对关系被保留，且尺寸范围不受time影响。
• 图层顺序：先画外层点，再画内层点，确保内层点始终显示在上方。
• 外层点样式：设置低透明度（alpha=0.3）和中性颜色（比如灰色），避免干扰内层点的视觉信息；用scale_size_identity(guide="none")隐藏外层点的尺寸图例，避免和内层点的图例冲突。

内容的提问来源于stack exchange，提问作者Ben