如何在ggplotly中移除箱线图异常值但保留抖动点

由于ggplotly会重新计算箱线图的统计指标，导致ggplot中outlier.shape = NA的设置失效，以下是几种可行的解决方法：

方法1：手动计算箱线图统计量（推荐） #

先手动计算箱线图所需的四分位数、上下须等统计量，再用stat="identity"绘制箱线图，让ggplotly直接沿用已计算好的数据，避免重新生成异常值点。

library(pacman)
pacman::p_load(dplyr, ggplot2, plotly)

# 计算箱线图统计量
box_stats <- chickwts %>%
  filter(feed == "sunflower") %>%
  summarize(
    ymin = quantile(weight, 0.25) - 1.5*IQR(weight),
    lower = quantile(weight, 0.25),
    middle = median(weight),
    upper = quantile(weight, 0.75),
    ymax = quantile(weight, 0.75) + 1.5*IQR(weight),
    feed = "sunflower"
  )

# 绘制箱线图+散点
g <- ggplot() +
  geom_boxplot(data = box_stats,
               aes(x = factor(feed), ymin = ymin, lower = lower, middle = middle, upper = upper, ymax = ymax, fill = factor(feed)),
               stat = "identity") +
  geom_jitter(data = chickwts %>% filter(feed == "sunflower"),
              aes(x = factor(feed), y = weight),
              alpha = 0.5, height = 0)

# 转换为ggplotly
ggplotly(g)

方法2：修改ggplotly的Trace参数 #

直接调整ggplotly生成的箱线图trace，将boxpoints参数设为"false"，强制隐藏异常值。

library(pacman)
pacman::p_load(dplyr, ggplot2, plotly)

g <- ggplot(chickwts %>% filter(feed == "sunflower"),
            aes(x = factor(feed), y = weight, fill = factor(feed))) + 
  geom_boxplot(outlier.shape = NA) + 
  geom_jitter(alpha = 0.5, height = 0)

# 调整trace参数隐藏异常值
ggplotly(g) %>%
  style(boxpoints = "false", traces = 1)

注：若存在多个箱线图trace，需根据实际情况调整traces的序号，或用which函数定位目标trace。

方法3：结合ggplot参数与ggplotly调整 #

在ggplot中设置outlier.alpha = 0隐藏异常点，再配合ggplotly的参数确保异常值不被渲染：

g <- ggplot(chickwts %>% filter(feed == "sunflower"),
            aes(x = factor(feed), y = weight, fill = factor(feed))) + 
  geom_boxplot(outlier.alpha = 0) + 
  geom_jitter(alpha = 0.5, height = 0)

ggplotly(g) %>% style(boxpoints = "false", traces = 1)

内容的提问来源于stack exchange，提问作者Majid Zaremehrjardy