R/QDC.Permute.R

In QDComparison: Modern Nonparametric Tools for Two-Sample Quantile and Distribution Comparisons

QDC.Permute <-
function (x, y, B = 500, m = 6, alpha = 0.05) 
{
    mat = matrix(NA, nrow = B, ncol = length(y))
    u_obs <- ecdf(y)(y)
    for (i in 1:B) {
        x_new <- sample(x)
        lp.boot.coef <- LP.XY(x_new, y, m, smooth = "FALSE")$LP.coef.1
        S_new <- eLP.poly(y, m)
        d.hat <- 1 + S_new %*% lp.boot.coef
        d.hat[d.hat < 0] <- 0
        u_new <- ecdf(y)(y)
        dfun <- approxfun(u_new, d.hat, rule = 2)
        mat[i, ] <- dfun(u_obs)
    }
    Q <- matrix(, 2, ncol(mat))
    for (j in 1:ncol(mat)) {
        Q[1, j] <- quantile(mat[, j], alpha/2)
        Q[2, j] <- quantile(mat[, j], 1 - alpha/2)
    }
    return(Q)
}