tests/testthat/test_postprocess.r

In simglm: Simulate Models Based on the Generalized Linear Model

context("test post process")

test_that('ifelse post process', {
  set.seed(321)

  sim_args <- list(
    formula = y ~ 1 +
      time +
      control +
      treat_knot +
      delay_knot +
      control_post +
      time:treat_knot +
      time:delay_knot +
      control:treat_knot +
      control:time:treat_knot +
      control_post:delay_knot +
      control_post:time:delay_knot +
      (1 + time | id),
    fixed = list(
      time = list(var_type = 'time', time_levels = seq(-5, 10, 1)),
      control = list(
        var_type = 'factor',
        levels = c('Treatment 2', 'Treatment 1'),
        var_level = 2
      )
    ),
    knot = list(
      treat_knot = list(variable = 'time', knot_locations = 0),
      delay_knot = list(variable = 'time', knot_locations = 4)
    ),
    post = list(
      control_post = list(
        variable = 'control',
        fun = 'ifelse',
        condition = '== "Treatment 2"',
        yes = 1,
        no = 0
      )
    ),
    sample_size = list(level1 = 16, level2 = 150),
    randomeffect = list(
      int_id = list(variance = 2, var_level = 2),
      time_id = list(variance = 0.5, var_level = 2)
    ),
    error = list(variance = 3),
    reg_weights = c(2, 0.5, 15, 0, 0, 0, 0, 0, 10, 8, 5, 3)
  )

  seg_data <- simulate_fixed(data = NULL, sim_args)

  expect_setequal(seg_data$control_1, seg_data$control_post)
  expect_setequal(seg_data$control_1 - seg_data$control_post, c(-1, 1))
  expect_equal(
    table(seg_data$control_post)[[1]],
    table(seg_data$control_1)[[2]]
  )
  expect_equal(
    table(seg_data$control_post)[[2]],
    table(seg_data$control_1)[[1]]
  )

  set.seed(1999)

  sim_arguments <- list(
    formula = retirement_savings ~ 1 +
      age +
      family_income +
      walkability +
      neighborhood_income_post +
      (1 | neighborhood),
    fixed = list(
      age = list(var_type = 'ordinal', levels = 20:80),
      family_income = list(var_type = 'continuous', mean = 0, sd = 25000),
      walkability = list(
        var_type = 'continuous',
        mean = 35,
        sd = 20,
        floor = 0,
        ceiling = 100
      )
    ),
    post = list(
      neighborhood_income_post = list(
        variable = 'family_income',
        by = 'neighborhood',
        fun = 'mean'
      )
    ),
    randomeffect = list(
      neighborhood_int = list(variance = 10000, var_level = 2)
    ),
    error = list(
      variance = 50000
    ),
    reg_weights = c(50000, 5000, 2, 1000, 0.5),
    sample_size = list(
      level1 = 1000,
      level2 = 50
    )
  )

  agg_data <- simulate_fixed(data = NULL, sim_arguments)

  expect_equal(length(unique(agg_data$neighborhood_income_post)), 50)
  expect_equal(
    mean(subset(agg_data, neighborhood == 1)$family_income),
    agg_data$neighborhood_income_post[1]
  )
})