Refactor: Extract helper functions to eliminate code duplication

[Copilot]

Identified and eliminated duplicated code patterns across 6 files, reducing ~160 lines of redundant logic.

Changes

Created R/utils.R

Extracted cross-file utilities:

• create_counterfactual_datasets() - Treatment counterfactual generation (used in gcomp_ml, gcomp_stan, mim)
• extract_treatment_coef_name() - Coefficient name extraction with factor level handling (used in calc_stc, mim)

Enhanced R/calculate_ate.R

Consolidated ALD data extraction and effect function dispatch:

• extract_ald_value() - Eliminates repeated dplyr::filter → pull chains across 6 trial calculation functions
• execute_effect_function() - Centralizes effect function validation and execution

Before:

calculate_trial_variance_binary <- function(ald, tid, effect) {
  y <- dplyr::filter(ald, .data$variable == "y", .data$trt == tid, 
                     .data$statistic == "sum") |> dplyr::pull(.data$value)
  N <- dplyr::filter(ald, .data$trt == tid, 
                     .data$statistic == "N") |> dplyr::pull(.data$value)
  
  effect_functions <- list(
    "log_odds" = function() 1/y + 1/(N-y),
    ...
  )
  
  if (!effect %in% names(effect_functions)) {
    stop(paste0("Unsupported effect function. Choose from ", ...))
  }
  effect_functions[[effect]]()
}

After:

calculate_trial_variance_binary <- function(ald, tid, effect) {
  y <- extract_ald_value(ald, tid, variable = "y", statistic = "sum")
  N <- extract_ald_value(ald, tid, variable = NULL, statistic = "N")
  
  effect_functions <- list(
    "log_odds" = function() 1/y + 1/(N-y),
    ...
  )
  
  execute_effect_function(effect, effect_functions)
}

Files Modified

• R/calculate_ate.R - 6 functions refactored (variance/mean calculations for binary, continuous, count data)
• R/gcomp_ml.R, R/gcomp_stan.R, R/mim.R - Counterfactual dataset creation
• R/calc_stc.R - Coefficient name extraction

All helper functions marked @keywords internal for package-internal use.

Original prompt

Find and refactor duplicated code

---

💬 We'd love your input! Share your thoughts on Copilot coding agent in our 2 minute survey.