I suppose the idea is that there is an underlying shape which the point cloud approximates. The fitted shape is then a statistical guess at the underlying shape

I suppose the idea is that there is an underlying shape which the point cloud approximates. The fitted shape is then a statistical guess at the underlying shape. In which case it would be nice to see a proof that as more points are added the fitted shape eventually converges to the underlying shape.