# Jackknife standard errors for estimates of gamma distribution parameters
# Inputs: vector of data values (data)
# Outputs: vector with two components, giving standard errors in shape and
  # scale parameters
gamma.jackknife <- function(data) {
  n <- length(data)
  # Prepare a matrix to store the re-estimates on partial data
  jackknifed.ests <- matrix(NA,nrow=2,ncol=n)
  rownames(jackknifed.ests) = c("a","s")
  # Omit each point in turn, and do estimation on the rest
  for (omitted.point in 1:n) {
    # Presumes gamma.est() exists in this environment and takes a data vector
    fit <- gamma.est(data[-omitted.point])
    # Presumes that gamma.est returns a list of two numbers named "a" and "s"
    jackknifed.ests["a",omitted.point] <- fit$a
    jackknifed.ests["s",omitted.point] <- fit$s
  }
  # Take the variance of each row of the re-estimates
  variance.of.ests <- apply(jackknifed.ests,1,var)
  # Scale up the variances to get the jackknife variance
  jackknifed.vars <- ((n-1)^2/n)*variance.of.ests
  # Standard error is the  square root of variance
  jackknifed.stderrs <- sqrt(jackknifed.vars)
  return(jackknifed.stderrs)
}