Discussion of Treelets by Lee, Nadler and Wasserman
Peter J. Bickel, Ya ' acov Ritov · 2007
The authors’ emphasis is on the method as a useful way of representing data analogous to a wavelet representation where X = X(t) with t genuinely identified with a point on the line and observartion at p time points ,but where the time points have been permuted. As such this can be viewed as a clustering method which, from their examples, gives very reasonable answers. However, to make more general theoretical statements and to permit comparison to other methods they necessarily introduce the model