def rand_err(mu, margins):
    sum = 0
    length = len(margins)
    # np.arange(margins.__len__())
    for i in np.arange(margins.__len__()):
        x = -1.0 * mu * margins[i]
        sum += pnorm(x)
    return sum / length