Local Extremes, Runs, Strings and Multiresolution

P. L. Davies, A. Kovac
2001 Annals of Statistics  
Laurie Davies and Arne Kovac have written a nice and very stimulating paper on nonparametric regression. They propose a novel approach to curve estimation, combining the traditional notion of estimation error (measured by supremum norm) and the complexity of the fitted function (measured by its modality). My comments are formulated in terms of a "true" underlying regression function f.
doi:10.1214/aos/996986501 fatcat:tyriac5dprdgbouag7bh7k2bdi