ggplot中为非绘图变量mpg与disp添加相关系数标签失败的问题

我绘制了如下散点图：

mtcars$mpg_level <- as.character(as.integer(mtcars$mpg))
p1 <- ggplot(data = mtcars, aes(x = disp, y = mpg_level, color=mpg_level , shape=mpg_level)) + geom_point()
p1 <- p1 + scale_shape_manual(values = 1:length(unique(mtcars$mpg_level)))

希望在图中添加mpg与disp的相关系数标签，但尝试以下代码时失败：

p2 <- ggplot(data = mtcars, aes(x = disp, y = mpg_level, color=mpg_level , shape=mpg_level)) + geom_point()
p2 <- p2 + stat_cor(aes(label = ..rr.label..), color = "red", geom = "label", digits = 2, size = 3)
p2 <- p2 + scale_shape_manual(values = 1:length(unique(mtcars$mpg_level)))

报错信息如下：

Warning message:
Computation failed in `stat_cor()`
Caused by error in `cor.test.default()`:
! not enough finite observations 

推测stat_cor试图计算每个mpg_level分组的相关系数，而直接以mpg为y轴时代码可以正常运行：

p3 <- ggplot(data = mtcars, aes(x = disp, y = mpg)) + geom_point()
p3 <- p3 + stat_cor(aes(label = ..rr.label..), color = "red", geom = "label", digits = 2, size = 3)

尝试单独为stat_cor指定x=disp和y=mpg，但该方法无效：

p4 <- ggplot(data = mtcars, aes(x = disp, y = mpg_level, color=mpg_level , shape=mpg_level)) + geom_point()
p4 <- p4 + stat_cor(aes(x=disp, y=mpg, label = ..rr.label..), color = "red", geom = "label", digits = 2, size = 3)
p4 <- p4 + scale_shape_manual(values = 1:length(unique(mtcars$mpg_level)))

请问如何在以mpg_level为y轴的散点图中添加mpg与disp的相关系数标签？

方法1：让stat_cor忽略分组映射 #

核心是给stat_cor指定group=1，强制它基于整个数据集计算相关系数，而非按mpg_level分组：

mtcars$mpg_level <- as.character(as.integer(mtcars$mpg))
p_final <- ggplot(data = mtcars, aes(x = disp, y = mpg_level, color=mpg_level , shape=mpg_level)) + 
  geom_point() +
  stat_cor(aes(x=disp, y=mpg, group=1, label = ..rr.label..), 
           color = "red", geom = "label", digits = 2, size = 3) +
  scale_shape_manual(values = 1:length(unique(mtcars$mpg_level)))
p_final

方法2：手动计算系数后用annotate添加标签 #

先手动计算mpg与disp的相关系数，再直接把标签标注到图上，位置可灵活调整：

# 手动计算相关系数（保留两位小数）
cor_r <- round(cor(mtcars$disp, mtcars$mpg), 2)
cor_label <- paste0("r = ", cor_r)

mtcars$mpg_level <- as.character(as.integer(mtcars$mpg))
p_final <- ggplot(data = mtcars, aes(x = disp, y = mpg_level, color=mpg_level , shape=mpg_level)) + 
  geom_point() +
  # 调整x、y参数可改变标签位置
  annotate("label", x = max(mtcars$disp)*0.8, y = max(mtcars$mpg_level), 
           label = cor_label, color = "red", size = 3) +
  scale_shape_manual(values = 1:length(unique(mtcars$mpg_level)))
p_final

内容的提问来源于stack exchange，提问作者curious