3.1 PKG

Show/Hide Code 
library(tidyverse)
library(hrbrthemes) # hrbrthemes 提供了更适合排版的主题
library(viridis) # viridis 提供了好看的色盲友好型颜色
library(ggdist) # ggdist 提供了半小提琴图和云雨图
library(patchwork) # patchwork 用于图形拼接
library(ggExtra) # ggExtra 用于 ggplot2 散点图的边际图

3.2 箱线图陷阱

箱线图又叫盒须图，展示数据的中位数(median)、上下四分位数(Quartiles)、四分位距(IQR)、须线(Whiskers)和异常值(outlier)。

这是说明箱线图构成的示意图： R for data science 对箱线图的介绍 r-graph-gallery 对箱线图的介绍

但是，r-graph-gallery 这个图的上下须比例不太对，我用 Base R 重新画了一个：

Show/Hide Code 
# --- 1. 参数设置 ---
# par(bg = "#eff8fc") # 背景

# 创建一个空白的绘图区域
plot(
    NA,
    xlim = c(-45, 145),
    ylim = c(2, 8),
    axes = FALSE,
    xlab = "",
    ylab = ""
)

# --- 2. 定义箱形图的统计值和位置 ---
y_center <- 5 # 图形在Y轴上的中心位置
box_height <- 2 # 箱体的高度
text_color <- "#000000" # 文本和箭头的颜色

# 定义关键的统计数值
q1 <- 35 # 第1四分位数 (Q1)
median_val <- 50 # 中位数 (Q2)
q3 <- 65 # 第3四分位数 (Q3)
iqr <- q3 - q1 # 四分位距 (IQR)

# 根据IQR定义须的末端位置（内限）
whisker_low <- q1 - 1.5 * iqr
whisker_high <- q3 + 1.5 * iqr

# 定义一些离群点 (outliers)
outliers_low <- whisker_low - c(8, 12, 15)
outliers_high <- whisker_high + c(8, 12, 15)

# --- 3. 绘制图形元素 ---
# 绘制箱体 (Box)
rect(
    q1,
    y_center - box_height / 2,
    q3,
    y_center + box_height / 2,
    border = "red",
    lwd = 3
)

# 绘制中位数线
segments(
    median_val,
    y_center - box_height / 2,
    median_val,
    y_center + box_height / 2,
    col = "red",
    lwd = 3
)

# 绘制上下须 (Whiskers)
segments(whisker_low, y_center, q1, y_center, col = "red", lwd = 3)
segments(q3, y_center, whisker_high, y_center, col = "red", lwd = 3)

# 绘制离群点
points(c(outliers_low, outliers_high), rep(y_center, 6), pch = 19, col = "red")

# --- 4. 添加注解和箭头 ---
# 中位数 (Median)
arrows(
    x0 = median_val,
    y0 = y_center + 2,
    x1 = median_val,
    y1 = y_center + box_height / 2 + 0.1,
    col = text_color,
    length = 0.1
)
text(
    x = median_val,
    y = y_center + 2.5,
    labels = "Median\n(Q2)",
    col = text_color
)

# Q1
arrows(
    x0 = q1 - 10,
    y0 = y_center + 2,
    x1 = q1,
    y1 = y_center + box_height / 2 + 0.1,
    col = text_color,
    length = 0.1
)
text(
    x = q1 - 10,
    y = y_center + 2.5,
    labels = "25th Percentile \n (Q1)",
    col = text_color
)

# Q3
arrows(
    x0 = q3 + 10,
    y0 = y_center + 2,
    x1 = q3,
    y1 = y_center + box_height / 2 + 0.1,
    col = text_color,
    length = 0.1
)
text(
    x = q3 + 10,
    y = y_center + 2.5,
    labels = "75th Percentile \n (Q3)",
    col = text_color
)

# 四分位距 (IQR)
segments(
    x0 = q1,
    y0 = y_center - 1.5,
    x1 = q3,
    y1 = y_center - 1.5,
    col = "red",
    lwd = 1.5
)
text(
    x = (q1 + q3) / 2,
    y = y_center - 2,
    labels = "Interquartile Range\n(IQR)",
    col = text_color
)

# 左侧离群点
arrows(
    x0 = mean(outliers_low) + 1,
    y0 = y_center - 1,
    x1 = outliers_low[2],
    y1 = y_center - 0.15,
    col = text_color,
    length = 0.1
)
arrows(
    x0 = mean(outliers_low) + 1,
    y0 = y_center - 1,
    x1 = outliers_low[1],
    y1 = y_center - 0.15,
    col = text_color,
    length = 0.1
)
arrows(
    x0 = mean(outliers_low) + 1,
    y0 = y_center - 1,
    x1 = outliers_low[3],
    y1 = y_center - 0.15,
    col = text_color,
    length = 0.1
)
text(
    mean(outliers_low) + 1,
    y_center - 1.3,
    "Potential\nOutliers",
    col = text_color
)

# 右侧离群点
arrows(
    x0 = mean(outliers_high) - 1,
    y0 = y_center - 1,
    x1 = outliers_high[2],
    y1 = y_center - 0.15,
    col = text_color,
    length = 0.1
)
arrows(
    x0 = mean(outliers_high) - 1,
    y0 = y_center - 1,
    x1 = outliers_high[1],
    y1 = y_center - 0.15,
    col = text_color,
    length = 0.1
)
arrows(
    x0 = mean(outliers_high) - 1,
    y0 = y_center - 1,
    x1 = outliers_high[3],
    y1 = y_center - 0.15,
    col = text_color,
    length = 0.1
)
text(
    mean(outliers_high) - 1,
    y_center - 1.3,
    "Potential\nOutliers",
    col = text_color
)

# 最小值标签 (Whisker end)
arrows(
    x0 = whisker_low,
    y0 = y_center - 2,
    x1 = whisker_low,
    y1 = y_center - 0.15,
    col = text_color,
    length = 0.1
)
text(
    x = whisker_low,
    y = y_center - 2.5,
    labels = "Minimum \n (Minimum Value in the Data, Q1 – 1.5*IQR)",
    col = text_color
)

# 最大值标签 (Whisker end)
arrows(
    x0 = whisker_high,
    y0 = y_center - 2,
    x1 = whisker_high,
    y1 = y_center - 0.15,
    col = text_color,
    length = 0.1
)
text(
    whisker_high,
    y_center - 2.5,
    "Maximum \n (Maximum Value in the Data, Q3 + 1.5*IQR)",
    col = text_color
)

# 数据中的真实最小值
arrows(
    x0 = min(outliers_low),
    y0 = y_center + 2,
    x1 = min(outliers_low),
    y1 = y_center + 0.15,
    col = text_color,
    length = 0.1
)
text(
    x = min(outliers_low),
    y = y_center + 2.5,
    labels = "Minimum Value \n in the Data",
    col = text_color
)

# 数据中的真实最大值
arrows(
    x0 = max(outliers_high),
    y0 = y_center + 2,
    x1 = max(outliers_high),
    y1 = y_center,
    col = text_color,
    length = 0.1
)
text(
    x = max(outliers_high),
    y = y_center + 2.5,
    labels = "Maximum Value \n in the Data",
    col = text_color
)

箱线图的构成 (由 Base R 绘制)

但是，这种信息的总结也有个大问题——无法显示数据的分布情况。例如：正态分布可能看起来与双峰分布完全相同。因此，考虑用小提琴图或脊线图。

Show/Hide Code 
# 创建数据集
data1 <- data.frame(
    name = c(
        rep("A", 500),
        rep("B", 250),
        rep("B", 250),
        rep("C", 20),
        rep("D", 100)
    ),
    value = c(
        rnorm(500, 10, 5),
        rnorm(250, 13, 1),
        rnorm(250, 18, 1),
        rnorm(20, 25, 4),
        rnorm(100, 12, 1)
    )
)

data1 |>
    ggplot(aes(x = name, y = value, fill = name)) +
    geom_boxplot() +
    scale_fill_viridis(discrete = TRUE) + # 好看的色盲友好型颜色,离散变量
    theme_ipsum() +
    theme(legend.position = "none") +
    labs(x = "") +
    ggtitle("A somewhat misleading boxplot")

3.3 改进

3.3.1 添加抖动(Jitter)

适合数据量不太大的情况

Show/Hide Code 
data1 |>
    ggplot(aes(x = name, y = value, fill = name)) +
    geom_boxplot() +
    scale_fill_viridis(discrete = TRUE) + # 好看的色盲友好型颜色,离散变量
    geom_jitter(color = "grey", size = 0.5, alpha = 0.5) +
    theme_ipsum() +
    theme(legend.position = "none") +
    labs(x = "") +
    ggtitle("A boxplot with jitter")

添加抖动(jitter)的箱线图

发现:

  • 组 C 样本量小。在得出组 C 的值高于其他组的结论之前，要考虑样本量.

  • 组 B 呈现出双峰分布(y = 18 和 y = 13),但是箱线图中看起来和组 A 并无区别.

3.3.2 小提琴图(Violin)

Show/Hide Code 
# 显示样本量
sample_size <- data1 |>
    group_by(name) |>
    summarize(num = n())

data1 |>
    left_join(sample_size) |>
    mutate(myaxis = paste0(name, "\n", "n=", num)) |>
    ggplot(aes(x = myaxis, y = value, fill = name)) +
    geom_violin(width = 1.4) +
    geom_boxplot(width = 0.1, color = "grey", alpha = 0.2) +
    scale_fill_viridis(discrete = TRUE) +
    theme_ipsum() +
    theme(legend.position = "none") +
    labs(x = "") +
    ggtitle("A violin plot")

3.3.3 云雨图(Raincloud)

看了就知道,云(半小提琴)+雨(散点)的组合。

Show/Hide Code 
data1 |>
    ggplot(aes(x = factor(name), y = value, fill = factor(name))) +

    # 添加半小提琴图（显示分布）
    stat_halfeye(
        adjust = 0.5,
        justification = -0.1,
        .width = 0,
        point_colour = NA
    ) +

    # 添加散点（显示原始数据点）
    stat_dots(
        side = "left",
        justification = 1.1,
        binwidth = 0.25
    ) +

    # 设置色盲友好型配色
    scale_fill_viridis(discrete = TRUE) +
    theme_ipsum() +
    theme(legend.position = "none") +
    labs(x = "") +
    ggtitle("A raincloud plot example")

把头顺时针旋转90度(或交换R代码X轴和Y轴)，就更像云雨了

甚至还可以再加上箱线图

Show/Hide Code 
data1 |>
    ggplot(aes(x = factor(name), y = value, fill = factor(name))) +

    # 添加半小提琴图（显示分布）
    stat_halfeye(
        adjust = 0.5,
        justification = -0.2,
        .width = 0,
        point_colour = NA
    ) +

    # 添加箱线图
    geom_boxplot(
        width = 0.12,
        outlier.color = NA,
        alpha = 0.5
    ) +

    # 添加散点（显示原始数据点）
    stat_dots(
        side = "left",
        justification = 1.1,
        binwidth = 0.25
    ) +

    # 设置色盲友好型配色
    scale_fill_viridis(discrete = TRUE) +
    theme_ipsum() +
    theme(legend.position = "none") +
    labs(x = "") +
    ggtitle("A raincloud plot example")

3.4 ggplot2

主要是geom_boxplot()函数.

3.4.1 基础

Show/Hide Code 
ggplot(mpg, aes(x = class, y = hwy)) +
    geom_boxplot()

最基础的箱线图
Show/Hide Code 
ggplot(mpg, aes(x = class, y = hwy)) +
    geom_boxplot(
        color = "blue", # 箱线图边框颜色为蓝色
        fill = "blue", # 箱体填充颜色为蓝色
        alpha = 0.2, # 箱体透明度为0.2，便于观察重叠部分

        notch = TRUE, # 显示缺口，用于比较中位数是否有显著差异
        notchwidth = 0.8, # 缺口的宽度

        outlier.colour = "red", # 异常值点的边框颜色为红色
        outlier.fill = "red", # 异常值点的填充颜色为红色
        outlier.size = 1 # 异常值点的大小为3
    )

有细腰的箱线图

3.4.2 排序

Show/Hide Code 
mpg |>
    # fct_reorder() 函数排序
    ggplot(aes(x = fct_reorder(class, hwy, .fun = "median"), y = hwy)) +
    geom_boxplot()

用均值排序箱线图

3.4.3 定制外观

Show/Hide Code 
p1 <- ggplot(mpg, aes(x = class, y = hwy)) +
    geom_boxplot(color = "red", fill = "orange", alpha = 0.2)

p2 <- ggplot(mpg, aes(x = class, y = hwy, fill = class)) +
    geom_boxplot(alpha = 0.3) +
    theme(legend.position = "none")

p3 <- ggplot(mpg, aes(x = class, y = hwy, fill = class)) +
    geom_boxplot(alpha = 0.3) +
    theme(legend.position = "none") +
    scale_fill_brewer(palette = "BuPu") # 调色板

p4 <- ggplot(mpg, aes(x = class, y = hwy, fill = class)) +
    geom_boxplot(alpha = 0.3) +
    theme(legend.position = "none") +
    scale_fill_brewer(palette = "Dark2") # 调色板

p1 + p2 + p3 + p4

改变颜色
Show/Hide Code 
mpg |>
    # 添加一列 'type'，用于标记是否高亮某个组
    mutate(type = ifelse(class == "subcompact", "Highlighted", "Normal")) |>
    ggplot(aes(x = class, y = hwy, fill = type, alpha = type)) +
    geom_boxplot() +
    scale_fill_manual(values = c("#69b3a2", "grey")) + # 手动设置填充色，高亮组为绿色，其余为灰色
    scale_alpha_manual(values = c(1, 0.1)) + # 手动设置透明度，高亮组为不透明，其余为半透明
    theme_ipsum() + # 使用 hrbrthemes 包的排版主题
    theme(legend.position = "none") + # 不显示图例
    xlab("") # 去除 x 轴标签

高亮某个组

3.4.4 分组/分面

Show/Hide Code 
# 构造数据
variety <- rep(LETTERS[1:7], each = 40) # 7种品种，每种40个观测
treatment <- rep(c("high", "low"), each = 20) # 处理分为high和low，每组20个观测
note <- seq(1:280) + sample(1:150, 280, replace = TRUE) # 生成note变量，添加一定随机性
data2 <- data.frame(variety, treatment, note) # 组合成数据框

# 分组箱线图
ggplot(data2, aes(x = variety, y = note, fill = treatment)) +
    geom_boxplot()

分组箱线图与分面箱线图
Show/Hide Code 
# 少分面箱线图
ggplot(data2, aes(x = variety, y = note, fill = treatment)) +
    geom_boxplot() +
    facet_wrap(~treatment)

分组箱线图与分面箱线图
Show/Hide Code 
# 多分面箱线图
ggplot(data2, aes(x = variety, y = note, fill = treatment)) +
    geom_boxplot() +
    facet_wrap(~variety, scale = "free") # 自由y轴

分组箱线图与分面箱线图

3.4.5 不等宽

让箱线图的宽度与样本量成正比

Show/Hide Code 
# 转换为因子类型
mpg$drv <- as.factor(mpg$drv)
# 创建x轴标签，包含每个drv水平的名称及其对应的样本量
n_xlab <- str_glue("{levels(mpg$drv)}\n(N={table(mpg$drv)})")

ggplot(mpg, aes(x = drv, y = hwy, fill = drv)) +
    geom_boxplot(varwidth = TRUE, alpha = 0.2) + # varwidth = TRUE 不等宽
    scale_x_discrete(labels = n_xlab) +
    theme(legend.position = "none")

不等宽箱线图(宽带是样本量)

3.4.6 连续变量

把连续变量分箱后再绘制箱线图。

Show/Hide Code 
diamonds |>
    mutate(bin = cut_width(carat, width = 0.5, boundary = 0)) |>
    ggplot(aes(x = bin, y = price)) +
    geom_boxplot()

连续变量箱线图

3.4.7 添加均值点

Show/Hide Code 
ggplot(mpg, aes(x = drv, y = hwy, fill = drv)) +
    geom_boxplot(alpha = 0.7) +
    stat_summary(fun = mean, geom = "point", shape = 1, size = 2, color = "red") +
    theme(legend.position = "none")

添加均值点

3.4.8 带数据点

Show/Hide Code 
# data1 是之前创建的数据集
data1 |>
    ggplot(aes(x = name, y = value, fill = name)) +
    geom_boxplot() +
    scale_fill_viridis(discrete = TRUE) + # 好看的色盲友好型颜色,离散变量
    geom_jitter(color = "grey", size = 0.5, alpha = 0.5) +
    theme_ipsum() + # 更适合排版的主题
    theme(legend.position = "none") +
    labs(x = "") +
    ggtitle("A boxplot with jitter")

带抖动的箱线图

3.4.9 外套箱线图

ggExtra包来实现更复杂（花哨）的图形，在ggplot2散点图的基础上再叠加箱线图、密度曲线等。

Show/Hide Code 
# 创建ggplot散点图
p <- ggplot(mtcars, aes(x = wt, y = mpg, color = cyl, size = cyl)) +
    geom_point() +
    theme(legend.position = "none")

ggMarginal(p, type = "histogram")

ggMarginal散点图叠加直方图
Show/Hide Code 
ggMarginal(p, type = "density")

ggMarginal散点图叠加密度图
Show/Hide Code 
ggMarginal(p, type = "boxplot")

ggMarginal散点图叠加箱线图

还可以定制化样式：

Show/Hide Code 
# 设置边际直方图的尺寸大小为10
ggMarginal(p, type = "histogram", size = 10)

ggMarginal定制化样式-尺寸大小
Show/Hide Code 
# 设置边际直方图的填充颜色为slateblue，x轴直方图分箱数为10
ggMarginal(p, type = "histogram", fill = "slateblue", xparams = list(bins = 10))

ggMarginal定制化样式-颜色和分箱
Show/Hide Code 
# 只在x轴添加边际图，边际图颜色为紫色，尺寸为4
ggMarginal(p, margins = "x", color = "purple", size = 4)

ggMarginal定制化样式-颜色和尺寸

3.5 Base R

主要是通过boxplot()函数.

但是 base R 多看一秒都是浪费时间,直接ggplot2吧.

如果实在想学,可以看 r-graph-gallery 的文档。

3.6 Pearl

Show/Hide Code 
# 自动安装packages
if (!require(ggstatsplot)) {
    install.packages(
        "ggstatsplot",
        repos = "https://cloud.r-project.org"
    )
}

if (!require(palmerpenguins)) {
    install.packages(
        "palmerpenguins",
        repos = "https://cloud.r-project.org"
    )
}

palmerpenguins::penguins |>
    drop_na() |>
    ggstatsplot::ggbetweenstats(x = species, y = flipper_length_mm, 1)

带有统计的小提琴箱线图

在 ggstatsplot 可以看到进一步美化。

或者

在 tidyplots 有另一种风格的统计箱线图。