tests/testthat/test_interactions.r

In simglm: Simulate Models Based on the Generalized Linear Model

context("Complex Interaction simulation")

test_that('3way interaction', {
  set.seed(1985)
  sim_arguments <- list(
    formula = y ~ 1 + sex + type + education + sex:type,
    fixed = list(sex = list(var_type = 'factor', levels=c('fem','male')),
                 type = list(var_type = 'factor', levels = c('type1', 'type2', 'type3')),
                 education = list(var_type = 'factor', 
                                  levels = c('freshman', 'sophomore', 'junior', 'senior'))),
    error = list(variance = 0.2),
    sample_size = 150,
    reg_weights = c(0,0,1,2,0, 0)    ## less parameters than previous example!
    # outcome_type = 'count'
  )
  fixed_vars <- simulate_fixed(data = NULL, sim_args = sim_arguments)
  
  expect_equal(ncol(fixed_vars), 13)
  expect_equal(length(unique(fixed_vars[['education']])), 4)
  expect_equal(length(unique(fixed_vars[['education_1']])), 2)
  expect_equal(length(unique(fixed_vars[['education_2']])), 2)
  expect_equal(length(unique(fixed_vars[['education_3']])), 2)
  
  sim_arguments <- list(
    formula = y ~ 1 + sex + type + education + sex:type + sex:education + education:type + sex:education:type,
    fixed = list(sex = list(var_type = 'factor', levels=c('fem','male')),
                 type = list(var_type = 'factor', levels = c('type1', 'type2', 'type3')),
                 education = list(var_type = 'factor', 
                                  levels = c('freshman', 'sophomore', 'junior', 'senior'))),
    error = list(variance = 0.2),
    sample_size = 150,
    reg_weights = c(0,0,1,2,0, 0)    ## less parameters than previous example!
    # outcome_type = 'count'
  )
  fixed_vars <- simulate_fixed(data = NULL, sim_args = sim_arguments)
  
  expect_equal(ncol(fixed_vars), 28)
  expect_equal(length(unique(fixed_vars[['education']])), 4)
  expect_equal(length(unique(fixed_vars[['education_1']])), 2)
  expect_equal(length(unique(fixed_vars[['education_2']])), 2)
  expect_equal(length(unique(fixed_vars[['education_3']])), 2)
  
  sim_arguments <- list(
    formula = y ~ 1 + sex + type + act + sex:type + sex:act + type:act + sex:act:type,
    fixed = list(sex = list(var_type = 'factor', levels=c('fem','male')),
                 type = list(var_type = 'factor', levels = c('type1', 'type2', 'type3')),
                 act = list(var_type = 'ordinal', 
                                  levels = 15:36)),
    error = list(variance = 0.2),
    sample_size = 150,
    reg_weights = c(0,0,1,2,0, 0)    ## less parameters than previous example!
    # outcome_type = 'count'
  )
  fixed_vars <- simulate_fixed(data = NULL, sim_args = sim_arguments)
  
  expect_equal(ncol(fixed_vars), 15)
  expect_type(fixed_vars[['act.sex_1.type_1']], 'double')
  expect_type(fixed_vars[['act.sex_1.type_2']], 'double')
  expect_type(fixed_vars[['type']], 'character')
})