R/confIntBootLogConROC_t0.r

In logcondens: Estimate a Log-Concave Probability Density from Iid Observations

confIntBootLogConROC_t0 <- function(controls, cases, grid = c(0.2, 0.8), conf.level = 0.95, M = 1000, smooth = TRUE, output = TRUE) {
  alpha <- 1 - conf.level

  boot.mat <- matrix(NA, nrow = length(grid), ncol = M)
  boot.mat.smooth <- boot.mat
  for (m in 1:M) {
    con.m <- sample(controls, replace = TRUE)
    cas.m <- sample(cases, replace = TRUE)
    roc <- logConROC(cas.m, con.m, grid, smooth = smooth)
    boot.mat[, m] <- roc$fROC
    if (identical(smooth, TRUE)) {
      boot.mat.smooth[, m] <- roc$fROC.smooth
    }
    if (identical(output, TRUE)) {
      print(paste(m, " of ", M, " runs done", sep = ""))
    }
  }

  ## quantiles of distribution at each point
  qs <- data.frame(cbind(grid, t(apply(boot.mat, 1, quantile, c(alpha / 2, 1 - alpha / 2)))))
  colnames(qs) <- c("t", "CIlow", "CIup")

  qs.smooth <- NA
  if (identical(smooth, TRUE)) {
    qs.smooth <- data.frame(cbind(grid, t(apply(boot.mat.smooth, 1, quantile, c(alpha / 2, 1 - alpha / 2)))))
    colnames(qs.smooth) <- c("t", "CIlow", "CIup")
  }

  ## generate output
  res <- list("quantiles" = qs, "boot.samples" = boot.mat, "quantiles.smooth" = qs.smooth, "boot.samples.smooth" = boot.mat.smooth)
  return(res)
}