R/indicators.est.R

In chiquadrat/ICD: Statistical Methods for Interval-Censored Data

# Internal documentation -------------------------------------------------------

# The function estimates the pre-defined indicators

indicators.est <- function(x, threshold, custom_indicator, weights) {
  if (is.null(weights)) {
    if (is.na(threshold)) {
      threshold <- 0.6 * median(x)
    } else {
      threshold <- threshold * median(x)
    }

    results <- c(
      mean = mean(x),
      gini = ineq(x),
      hcr = mean(x < threshold),
      quant10 = unname(quantile(x, probs = c(0.1))),
      quant25 = unname(quantile(x, probs = c(0.25))),
      quant50 = unname(quantile(x, probs = c(0.50))),
      quant75 = unname(quantile(x, probs = c(0.75))),
      quant90 = unname(quantile(x, probs = c(0.90))),
      pgap = mean((x < threshold) * (threshold - x) / threshold),
      qsr = sum(x[(x > quantile(x, 0.8))]) / sum(x[(x < quantile(x, 0.2))])
    )

    if (!is.null(custom_indicator)) {
      c.indicator <- NULL
      for (i in 1:length(custom_indicator)) {
        c.indicator <- c(c.indicator, custom_indicator[[i]](x, threshold))
      }
      names(c.indicator) <- names(custom_indicator)
      results <- c(results, c.indicator)
    }
  } else {
    sw <- sum(weights)
    if (is.na(threshold)) {
      thresholdW <- 0.6 * wtd.quantile(x,
        weights = weights,
        probs = .5
      )
    } else {
      thresholdW <- threshold * wtd.quantile(x,
        weights = weights,
        probs = .5
      )
    }

    results <- c(
      mean = weighted.mean(x, w = weights),
      gini = gini(x, weights = weights)$value / 100,
      hcr = arpr(x, weights = weights, p = threshold)$value / 100,
      quant10 = unname(wtd.quantile(x, weights = weights, probs = 0.1)),
      quant25 = unname(wtd.quantile(x, weights = weights, probs = 0.25)),
      quant50 = unname(wtd.quantile(x, weights = weights, probs = 0.5)),
      quant75 = unname(wtd.quantile(x, weights = weights, probs = 0.75)),
      quant90 = unname(wtd.quantile(x, weights = weights, probs = 0.9)),
      pgap = sum(weights * (x < thresholdW) * (thresholdW - x)
        / thresholdW) / sw,
      qsr = qsr(x, weights = weights)$value
    )

    if (!is.null(custom_indicator)) {
      c.indicator <- NULL
      for (i in 1:length(custom_indicator)) {
        c.indicator <- c(c.indicator, custom_indicator[[i]](x, threshold, weights))
      }
      names(c.indicator) <- names(custom_indicator)
      results <- c(results, c.indicator)
    }
  }

  return(results)
}