# Card (1993) Figure 1 – mean education by predicted-education quartile × nearc4.
library(tidyverse)
library(here)
library(ggplot2)
source(here("04-topics/rep-card1993/Rcode/card1993-data-prep.R"))

data <- load_card93_data()

# Age dummies (1966 age groups, same bins as Table 1).
data <- data |>
  mutate(
    age_1415 = as.integer(age66 >= 14 & age66 <= 15),
    age_1617 = as.integer(age66 >= 16 & age66 <= 17),
    age_1820 = as.integer(age66 >= 18 & age66 <= 20),
    age_2124 = as.integer(age66 >= 21 & age66 <= 24)
  )

pred_x <- c(
  card93_region66, "smsa66r",
  "age_1415", "age_1617", "age_1820", "age_2124",
  "black", card93_fb_full
)

no_college <- data |> filter(nearc4 == 0L)
fml <- as.formula(paste("ed76 ~", paste(pred_x, collapse = " + ")))
fit <- lm(fml, data = no_college)

data$pred_ed_nocollege <- predict(fit, newdata = data)
data$pred_quartile <- ntile(data$pred_ed_nocollege, 4L)

fig_summary <- data |>
  group_by(pred_quartile, nearc4) |>
  summarise(mean_ed = mean(ed76), n = n(), .groups = "drop") |>
  mutate(
    nearc4_lab = if_else(nearc4 == 1L, "Near 4-year college", "No nearby college")
  )

fig_plot <- fig_summary |>
  ggplot(aes(x = pred_quartile, y = mean_ed, colour = nearc4_lab, group = nearc4_lab)) +
  geom_line(linewidth = 0.9) +
  geom_point(size = 2.5) +
  scale_x_continuous(breaks = 1:4, labels = paste("Q", 1:4)) +
  scale_colour_manual(values = c("No nearby college" = "#2166ac", "Near 4-year college" = "#b2182b")) +
  labs(
    x = "Quartile of predicted education (no college nearby subsample fit)",
    y = "Mean completed education in 1976 (years)",
    colour = NULL,
    title = "Mean completed education by predicted-education quartile and college proximity in 1966",
    caption = "Note: Prediction equation is fit to subsample with no college nearby."
  ) +
  theme_minimal(base_size = 12) +
  theme(legend.position = "bottom")

r2_nocollege <- summary(fit)$r.squared

paper_note <- tibble(
  metric = c("R2 prediction (no college subsample)", "Lowest quartile gap (near - no)"),
  value = c(
    r2_nocollege,
    {
      q1 <- fig_summary |> filter(pred_quartile == 1L)
      q1$mean_ed[q1$nearc4 == 1L] - q1$mean_ed[q1$nearc4 == 0L]
    }
  )
)

message("Figure 1 summary:")
print(fig_summary)
print(paper_note)

save(fig_plot, fig_summary, paper_note, fit, file = here("04-topics/rep-card1993/Rcode/Figure_1.RData"))