## The sample sizes.
n = 10
m = 10

## Generate permutation-test p-values for 1000 null data sets.
pv = array(0, 1000)
for (j in (1:1000))
{
  ## Variances for the two populations.
  V = rexp(2)

  ## Generate the 'actual' data.
  X = sqrt(V[1])*rnorm(n)
  Y = sqrt(V[2])*rnorm(m)

  ## Calculate the Z-statistic for the actual data.
  mx = mean(X)
  my = mean(Y)
  vx = var(X)
  vy = var(Y)
  T = (mx - my) / sqrt(vx/n + vy/n)

  ## Merge all the data together.
  Z = c(X, Y)

  ## Get 1000 Z-statistics for permuted data.
  TR = array(0, 1000)
  for (r in (1:1000))
  {
    ## Generate a random permutation.
    ii = order(runif(m+n))

    ## Construct x and y data sets by random reassignment.
    x = Z[ii[1:n]]
    y = Z[ii[(n+1):(n+m)]]

    ## Calculate the Z-statistic for the reassigned data.
    mx = mean(x)
    my = mean(y)
    vx = var(x)
    vy = var(y)
    TR[r] = (mx - my) / sqrt(vx/n + vy/n)
  }

  ## Two-sided p-value.
  pv[j] = mean(abs(TR) > abs(T))
}