预测的回归线与 geom_smooth(method= "lm", formula= "y ~ x") 中的不一样

正如用户20650已经在 comments 中指出的那样，ggplot添加了一个交互术语，而您的模型mod只考虑了基于附加变量+ group的组差异.要获得相同的结果，要么使用交互术语* group，要么 for each 组创建两个不同的模型.请参阅下面的两种方法:

# Make some data
set.seed(1)
n <- 100
df <- data.frame(x= rnorm(n, 100, 15), group= sample(0:1, n/2, replace= TRUE))
df$y <- df$x*.1 + df$group*2 + rnorm(n)

# Fit data and predict values

# With interaction
mod <- lm(y ~ x * group, df) # <- add an interaction term
df_new <- setNames(data.frame(expand.grid(round(seq(min(df$x), max(df$x), 1), 0),
                                          0:1)),
                   nm= c("x", "group"))
df_pred <- cbind.data.frame(df_new, pred = predict(mod, df_new))


# Predict for each group
mod0 <- lm(y ~ x, df[df$group == 0, ]) 

df_new0 <- setNames(data.frame(expand.grid(round(seq(min(df$x), max(df$x), 1), 0),
                                          0)),
                   nm= c("x", "group"))


df_pred0 <- cbind.data.frame(df_new0, pred = predict(mod0, df_new0))

mod1 <- lm(y ~ x, df[df$group == 1, ])

df_new1 <- setNames(data.frame(expand.grid(round(seq(min(df$x), max(df$x), 1), 0),
                                           1)),
                    nm= c("x", "group"))

df_pred1 <- cbind.data.frame(df_new1, pred = predict(mod1, df_new1))

df_pred2 <- rbind(df_pred0, df_pred1)

# Plot data with predicted values and geom_smooth()
library(ggplot2)

ggplot(df, aes(x, y, col = factor(group))) +
  geom_point() +
  geom_smooth(method= "lm", formula= "y ~ x") +
  geom_line(data=df_pred, mapping= aes(x, pred, col = factor(group))) + # with interaction
  geom_line(data=df_pred2, mapping= aes(x, pred, col = factor(group))) # with seperate groups