在 ggplot 中用文本和框（和箭头）标记特定数据点

Question

我有一个散点图和基础回归模型。我想在另一个数据点上给出一个很好的例子（假设不包含在估计样本中），即它的实际值与预测值。除了标签之外，我已经准备好了一切（考虑geom_text或geom_label，到目前为止没有任何效果）：

data(mtcars)

model <- lm(mpg ~ wt, data=mtcars)
mycar <- data.frame(wt=c(2.5))
predict(model, mycar)

model <- lm(Coupon ~ Total, data=df)
mycar <- data.frame(Total=c(79037022, 83100656, 104299800))
predict(model, new.df)

ggplot(mtcars, aes(x=wt, y=mpg)) +
  geom_point() +
  geom_smooth(method="lm", se=FALSE) +
  geom_point(aes(x=2.5,y=23.92395),   # red is the prediction for my car
             color='red',
             size=3, show.legend = TRUE) + 
  geom_point(aes(x=2.5,y=28),         # green is the actual mpg of my car
             color='green',
             size=3) 

我对具体实施很灵活：

• 文本的颜色可以与数据点的颜色相同（没有线条或箭头）
• 带线条或箭头的黑色文本（需要框背景，因为在实际数据中，数据云几乎到处都是黑色）
• 可能有一个图例将手动添加点的颜色映射到文本标签（尽管我喜欢数据点旁边的标签）。
• ...

我正在寻找易于实施且具有视觉吸引力的东西。谢谢

Answer 1

我推荐该ggrepel包提供一些功能，以避免与绘图上已绘制的现有项目重叠，特别是当您有多个标签/文本时

library(ggplot2)
library(ggrepel)
data(mtcars)

model <- lm(mpg ~ wt, data = mtcars)
mycar <- data.frame(wt = c(2.5))

ggplot() +
  geom_point(data = mtcars, aes(x = wt, y = mpg)) +
  geom_smooth(data = mtcars, aes(x = wt, y = mpg), method = "lm", se = FALSE) +
  geom_point(aes(x = 2.5, y = 23.92395), # red is the prediction for my car
    color = "red",
    size = 3, show.legend = TRUE
  ) +
  geom_point(aes(x = 2.5, y = 28), # green is the actual mpg of my car
    color = "green",
    size = 3
  ) +
  geom_label_repel(aes(x = 2.5, y = 23.92395),
    # using nudge_x & nudge_y to manual adjust the location to avoid
    # label overlay on a data point around 3.5 & 25    
    color = "red", label = "My car prediction", nudge_x = .5, nudge_y = 1.5
  ) +
  geom_label_repel(aes(x = 2.5, y = 28),
    color = "green", label = "My actual car", nudge_x = .5, nudge_y = 1
  )
#> `geom_smooth()` using formula 'y ~ x'

ggplot() +
  geom_point(data = mtcars, aes(x = wt, y = mpg)) +
  geom_smooth(data = mtcars, aes(x = wt, y = mpg), method = "lm", se = FALSE) +
  geom_point(aes(x = 2.5, y = 23.92395), # red is the prediction for my car
    color = "red",
    size = 3, show.legend = TRUE
  ) +
  geom_point(aes(x = 2.5, y = 28), # green is the actual mpg of my car
    color = "green",
    size = 3
  ) +
  geom_text_repel(aes(x = 2.5, y = 23.92395),
    color = "red", label = "My car prediction", nudge_x = .5, nudge_y = 1
  ) +
  geom_text_repel(aes(x = 2.5, y = 28),
    color = "green", label = "My actual car", nudge_x = .5, nudge_y = 1
  )
#> `geom_smooth()` using formula 'y ~ x'

^{由reprex 包(v2.0.0)创建于 2021-05-04}