如何获取多组线性回归的斜率并在ggplot分面图中标注？

一、提取各组回归斜率 #

按quantity（分面变量）和quality（分组变量）分别拟合线性模型，提取斜率值。可以用dplyr结合broom包高效完成：

library(dplyr)
library(broom)
library(ggplot2)

# 分组拟合模型并提取斜率
slope_results <- donnees_tot %>%
  group_by(quantity, quality) %>%
  do(model_fit = lm(weight ~ temperature, data = .)) %>%
  mutate(slope = tidy(model_fit)$estimate[tidy(model_fit)$term == "temperature"]) %>%
  select(quantity, quality, slope) %>%
  ungroup()

# 查看结果
print(slope_results)

二、将斜率（或完整回归方程）标注到图中 #

修改你的ggplot代码，加入geom_text（或geom_label），利用生成的斜率数据，将标注放在每个分面的合适位置：

1. 仅标注斜率值

g = ggplot(donnees_tot, aes(x=temperature, y=weight, col = quality))+
  geom_point(size = 3)+
  geom_smooth(method="lm", aes(col=quality, fill=quality))+ # lm方法下span参数无效，已移除
  scale_color_manual(values=c("S" = "aquamarine3", "Y" = "darkgoldenrod3"))+
  scale_fill_manual(values=c("S" = "aquamarine3", "Y" = "darkgoldenrod3"))+
  scale_x_continuous(breaks=c(20,25,28), limits=c(20,28))+
  # 添加斜率标注，调整x/y坐标到合适位置
  geom_text(data = slope_results,
            aes(x = 27, y = max(donnees_tot$weight, na.rm = TRUE)*0.9,
                label = paste("斜率 =", round(slope, 2)),
                col = quality),
            hjust = 1)+
  facet_wrap(~quantity)
g

2. 标注完整回归方程（支持公式解析）

如果需要显示类似y = ax + b的公式，先生成公式文本，再用parse = TRUE解析：

# 生成包含截距和斜率的回归方程文本
equation_results <- donnees_tot %>%
  group_by(quantity, quality) %>%
  do(model_fit = lm(weight ~ temperature, data = .)) %>%
  mutate(
    intercept = tidy(model_fit)$estimate[tidy(model_fit)$term == "(Intercept)"],
    slope = tidy(model_fit)$estimate[tidy(model_fit)$term == "temperature"],
    # 生成ggplot可解析的公式文本
    eqn = paste0("weight == ", round(intercept, 2), " + ", round(slope, 2), "*temperature")
  ) %>%
  select(quantity, quality, eqn) %>%
  ungroup()

# 绘图并添加公式标注
g = ggplot(donnees_tot, aes(x=temperature, y=weight, col = quality))+
  geom_point(size = 3)+
  geom_smooth(method="lm", aes(col=quality, fill=quality))+
  scale_color_manual(values=c("S" = "aquamarine3", "Y" = "darkgoldenrod3"))+
  scale_fill_manual(values=c("S" = "aquamarine3", "Y" = "darkgoldenrod3"))+
  scale_x_continuous(breaks=c(20,25,28), limits=c(20,28))+
  geom_text(data = equation_results,
            aes(x = 27, y = max(donnees_tot$weight, na.rm = TRUE)*0.9,
                label = eqn,
                col = quality),
            hjust = 1, parse = TRUE)+
  facet_wrap(~quantity)
g

注意事项 #

• span参数是给loess平滑方法用的，在method="lm"时不会生效，可直接移除。
• 标注的x和y坐标可根据数据范围调整，比如把y设为每组数据的最大值附近，避免遮挡数据点。

内容的提问来源于stack exchange，提问作者Nate Trf