R/kenwardroger.R

In meta: General Package for Meta-Analysis

kenwardroger <- function(w.random) {
  ##
  ## Kenward-Roger method for meta-analysis
  ## (Partlett & Riley, 2017, Stat Med)
  ##
  w1p <- sum(w.random)
  w2p <- sum(w.random^2)
  w3p <- sum(w.random^3)
  ##
  IE <- w2p * 0.5 - w3p / w1p + 0.5 * (w2p / w1p)^2
  var <- (1 + 2 * (w3p / w1p - (w2p / w1p)^2) / IE) / w1p
  df <- 2 * IE / (var * w2p)^2
  ##
  res <- list(se = sqrt(var), df = df,
              IE = IE, w1p = w1p, w2p = w2p, w3p = w3p)
  res
}