tests/testthat/test-bootstrap-multiple-variables.R

In BSW: Fitting a Log-Binomial Model Using the Bekhit–Schöpe–Wagenpfeil (BSW) Algorithm

test_that("Bootstrap process runs successfully with multiple variables", {

  # Create example data
  set.seed(123)

  # Generate 5 independent variables from normal distributions
  x1 <- rnorm(1000, 50, 10)
  x2 <- rnorm(1000, 30, 5)
  x3 <- rnorm(1000, 40, 8)
  x4 <- rnorm(1000, 60, 12)
  x5 <- rnorm(1000, 55, 7)

  # Compute the dependent variable y using a logistic transformation
  logit <- -4 + x1 * 0.04 + x2 * 0.05 + x3 * 0.02 + x4 * -0.03 + x5 * 0.01
  p <- 1 / (1 + exp(-logit))
  y <- rbinom(1000, 1, p)

  data <- data.frame(y, x1, x2, x3, x4, x5)

  # Fit the model with a subset of predictors
  fit <- bsw(formula = y ~ x1 + x2 + x3, data = data)

  # Run the bootstrap procedure
  result <- bootbsw(fit, ci_level = 0.90, R = 1000L)

  # Extract the bootstrap object
  bootobj <- result$Bootstrap_Object

  # Checks
  expect_true(is.numeric(bootobj$t0))
  expect_equal(length(bootobj$t[,1]), 1000)
  expect_gt(sum(complete.cases(bootobj$t)), 950)
})