Search for the QCD Ground State.pdf

Search of Signicant Variables in Nonparametric Additive Regression W Hardle Institut fur Statistik und Okonometrie Wirtschaftswissenschaftliche Fakultat Humb oldtUniversitat zu Berlin Spandauer Strae D Berlin Germany A Korostelev Institute for System Analysis Prosp ekt Let Oktjabrja Moscow Russia Abstract Nonparametric additive regression is studied under the assumption that only a part of nonparametric comp onents is separated away from zero Each of these nonzero comp onents dep ends on its own particular explanatory variable called a signicant variable The search problem for signicant variables is considered and the algorithm is prop osed which guarantees the exp onentially fast decreasing error probabilities as the sample size grows  The research desrib ed in this pap er was made p ossible in part by Grant MCF from the Interna tional Scientic Foundation The research for this pap er was carried out within Sonderforschungsb e reich at the Humb oldtUniversity Berlin and was printed using funds made available by the Deutsche Forschungsgemeinschaft Supp ort of INRA INSEE Paris France is gratefully acknowlegded INTR ODUCTION Intro duction Consider a nonparametric regression mo del with a ddimensional explanatory variable X X (1) X (d) and a onedime