## Generate nrep datasets from an AR(1) process.
arsim = function(alpha, t2, nrep, n) {

    ## The data variance.
    s2 = t2 / (1 - alpha^2)

    ## Storage for nrep dependent sequences of length n, each stored
    ## in a row of X.
    X = array(0, c(nrep,n))

    ## Simulate the first value.
    X[,1] = rnorm(nrep, sd=sqrt(s2))

    ## Simulate the rest of the sequence.
    for (i in (2:n))
    {
        X[,i] = alpha*X[,i-1] + rnorm(nrep, sd=sqrt(t2))
    }

    return(X)
}