The elimination of underestimation in nearest-neighbour analysis