Remove birthweight dataset references and fix 4 test failures

[Copilot]

The birthweight dataset was replaced by cachar_sample but leftover test references and several unrelated test bugs were causing 10 CI failures.

Changes

• Deleted obsolete birthweight test (test-calc-risk-diff.R): Removed test_that("calc_risk_diff works with package birthweight data", ...) entirely — dataset no longer exists in the package.

• Fixed create_convergence_challenge_data() (test-calc-risk-diff.R): data.frame() evaluates arguments in the calling environment, not incrementally, so exposure was not in scope when computing outcome_prob. Extracted both as local variables before the data.frame() call:

  # Before (broken): exposure not in scope inside data.frame()
  data.frame(
    exposure = factor(c(rep("No", n*0.8), rep("Yes", n*0.2))),
    outcome_prob = ifelse(exposure == "Yes", 0.85, 0.05),  # error: object 'exposure' not found
    ...
  )
  
  # After: pre-compute outside data.frame()
  exposure <- factor(c(rep("No", n*0.8), rep("Yes", n*0.2)))
  outcome_prob <- ifelse(exposure == "Yes", 0.85, 0.05)
  data.frame(exposure = exposure, outcome = rbinom(n, 1, outcome_prob), ...)

• Fixed boundary detection SE test (test-boundary-detection-enhanced.R): n=50 with 1 confounder wasn't pushing SE/coefficient ratio above the detection threshold. Reduced to n=30 and added confounders x1–x4 to overparameterize the model and reliably trigger large_standard_errors.

• Fixed sex-stratified result check (test-calc-risk-diff.R): When a stratum row is missing, male_result$rd returns numeric(0), making if (!is.na(...)) throw "argument is of length zero". Added nrow() > 0 guard:

  if (nrow(male_result) > 0 && !is.na(male_result$rd)) expect_true(male_result$rd > 0)

Original prompt

Remove All Birthweight Dataset References and Fix Remaining Test Failures

The package previously contained a birthweight dataset that was replaced with cachar_sample synthetic data. However, leftover references to birthweight remain in the codebase and are causing test failures.

10 Test Failures to Fix:

Primary Issues:

1. Birthweight dataset reference (test-calc-risk-diff.R:636)

   • Test: "calc_risk_diff works with package birthweight data"
   • Error: object 'birthweight' not found
   • Solution: Delete this entire test block (lines ~631-651) - it's obsolete

2. create_convergence_challenge_data() has a variable reference error (test-calc-risk-diff.R:400)

   • Error: object 'exposure' not found
   • Problem: Function tries to use exposure in a piped operation before it's created in the dataframe
   • Solution: Use proper column reference syntax within mutate() or create exposure first

3. Large standard errors boundary test not triggering detection (test-boundary-detection-enhanced.R:103)

   • Test: "boundary detection handles large standard errors"
   • Failure: boundary_large_se$boundary_type doesn't match expected values
   • Solution: Add more confounders (x1, x2, x3, x4) and use smaller n to push SE/coef ratio > 2

4. Sex-stratified results returning empty dataframes (test-calc-risk-diff.R:606)

   • Test: "calc_risk_diff shows expected sex differences in tobacco use patterns"
   • Error: argument is of length zero when accessing male_result$rd
   • Solution: Add defensive check for empty results: if (nrow(male_result) > 0) { ... }

All References to Remove:

Search across entire repo for "birthweight" and replace/remove:

• Test files
• Documentation files
• Comments
• R data files
• Examples

Expected Outcome:

• All 10 test failures resolved
• 0 test failures, 588 passing tests
• All birthweight references completely removed
• Package passes R CMD check without errors

This pull request was created from Copilot chat.

---

💡 You can make Copilot smarter by setting up custom instructions, customizing its development environment and configuring Model Context Protocol (MCP) servers. Learn more Copilot coding agent tips in the docs.