R/6_Main.R

In zrmacc/MargRates: Marginal Event Rates

#' Compare Marginal Event Rates.
#' 
#' Given stratified event count data for two arms, analyze the marginal
#' event rates via the risk difference, risk ratio, and odds ratio.
#' 
#' @param y0 Events per category in arm 0.
#' @param n0 Subjects per category in arm 0.
#' @param y1 Events per category in arm 1.
#' @param n1 Subjects per category in arm 1.
#' @param weights Stratum mixing weights. If omitted, defaults to the 
#'   stratum proportions.
#' @param alpha Type I error. 
#' @param boot Calculate bootstrap confidence intervals? 
#' @param perm Perform a permutation-type test of the null?
#' @param reps Replicates for bootstrap/permutation.
#' @param exclude_double_zero Exclude strata with no events in either arm?
#' @importFrom methods new
#' @export
#' @return Object of class `margRates` containing these slots:
#' \itemize{
#'   \item `@Rates`, the marginal event rates in each arm.
#'   \item `@RD`, risk difference analysis.
#'   \item `@RR`, risk ratio analysis.
#'   \item `@OR`, odds ratio analysis.
#' }

CompMargRates <- function(
  y0, 
  n0, 
  y1, 
  n1, 
  weights = NULL,
  alpha = 0.05,
  boot = FALSE,
  perm = FALSE,
  reps = 2e3,
  exclude_double_zero = FALSE
) {
  
  # Double zeros.
  if (exclude_double_zero) {
    is_double_zero <- (y0 == 0) & (y1 == 0)
    y0 <- y0[!is_double_zero]
    y1 <- y1[!is_double_zero]
    n0 <- n0[!is_double_zero]
    n1 <- n1[!is_double_zero]
  }
  
  # Weights.
  if (!is.null(weights)) {
    weights <- weights / sum(weights)
  }
  
  # Asymptotic inference.
  asymp <- CalcMargStats(
    y0 = y0,
    n0 = n0,
    y1 = y1,
    n1 = n1,
    weights = weights,
    alpha = alpha
  )
  
  # Rates.
  rates <- asymp$Rates
  stats_asymp <- asymp$Stats
  stats_asymp$Method <- "Asymptotic"
  
  # Bootstrap inference.
  if (boot) {
    stats_boot <- Stats.Boot(
      y0 = y0,
      n0 = n0,
      y1 = y1,
      n1 = n1,
      weights = weights,
      alpha = alpha,
      reps = reps
    )
    stats_boot$Method <- "Bootstrap"
  } else {
    stats_boot <- NULL
  }
  
  # Permutation inference.
  if (perm) {
    perm_test <- Test.Null(
      y0 = y0,
      n0 = n0,
      y1 = y1,
      n1 = n1,
      weights = weights,
      alpha = alpha,
      reps = reps
    )
  } else {
    perm_test <- data.frame()
  }
  
  # Output.
  out <- rbind(stats_asymp, stats_boot)
  out <- out[, c(7, 1:6)]
  out <- new(
    Class = "margRates",
    Rates = rates,
    RD = out[out$Stat == "RiskDiff", ],
    RR = out[out$Stat == "RiskRatio", ],
    OR = out[out$Stat == "OddsRatio", ],
    Perm = perm_test
  )
  return(out)
}