# this script creates the dataset for the multilevel exercise

library(dplyr)
library(tidyr)
library(faux)
library(lme4)

set.seed(123)
n1 <- 20
n2 <- 15
dat <-
  rnorm_multi(n = n2,
              mu = 0:1,
              sd = 1,
              r = .5,
              varnames = paste0("u", 0:1, "j")) |> 
  mutate(j = 1:n2,
         n1 = n1) |> 
  uncount(n1) |> 
  bind_cols(rnorm_multi(n = n1*n2,
                        r = .5,
                        varnames = paste0("x", 1:3, "i"))) |> 
  mutate(y = u0j + x1i + x2i + u1j*x3i + rnorm(n1*n2))

lmer(y ~ x1i + x2i + x3i + (x3i|j), dat) |> summary()

dat <-
  dat |> 
  mutate(
    x1i = ifelse((x1i > quantile(x1i, .4)) & (runif(n1*n2) > .5), NA, x1i), 
    x2i = ifelse((x3i > quantile(x3i, .5)) & (runif(n1*n2) > .5), NA, x2i)
  )

lmer(y ~ x1i + x2i + x3i + (x3i|j), dat) |> summary()

# write.csv(dat |> select(!starts_with("u")), "exercise_04.csv", row.names = F)