example_of_cross_validation

x <- rnorm(1000, 7, 2)
noise <- rnorm(1000)

y <- 5 + 2 * x + noise

df <- data.frame(x = x, y = y)

test_idx <- sample(1000, 100)

test_df <- df[test_idx, ]

train_df <- df[!(seq_len(1000) %in% test_idx), ]

my_model <- lm(y ~ x, data = train_df)

my_estimate <- test_df %>% 
    mutate(est_val = predict(my_model, newdata = .),
           res = est_val - y)


ggplot(my_estimate, aes(x, y = res)) +
    geom_point() +
    geom_hline(yintercept = mean(my_estimate$res), color = "red")