Variable Bandwidth Kernel Estimators of Regression Curves
Muller, Hans-Georg ; Stadtmuller, Ulrich
Ann. Statist., Tome 15 (1987) no. 1, p. 182-201 / Harvested from Project Euclid
In the model $Y_i = g(t_i) + \varepsilon_i,\quad i = 1,\cdots, n,$ where $Y_i$ are given observations, $\varepsilon_i$ i.i.d. noise variables and $t_i$ nonrandom design points, kernel estimators for the regression function $g(t)$ with variable bandwidth (smoothing parameter) depending on $t$ are proposed. It is shown that in terms of asymptotic integrated mean squared error, kernel estimators with such a local bandwidth choice are superior to the ordinary kernel estimators with global bandwidth choice if optimal bandwidths are used. This superiority is maintained in a certain sense if optimal local bandwidths are estimated in a consistent manner from the data, which is proved by a tightness argument. The finite sample behavior of a specific local bandwidth selection procedure based on the Rice criterion for global bandwidth choice [Rice (1984)] is investigated by simulation.
Publié le : 1987-03-14
Classification:  Nonparametric kernel regression,  consistent bandwidth choice,  Rice criterion,  local bandwidths,  asymptotic optimality,  tightness in $C$,  Gaussian limiting process,  62G05,  65J02,  65D10
@article{1176350260,
     author = {Muller, Hans-Georg and Stadtmuller, Ulrich},
     title = {Variable Bandwidth Kernel Estimators of Regression Curves},
     journal = {Ann. Statist.},
     volume = {15},
     number = {1},
     year = {1987},
     pages = { 182-201},
     language = {en},
     url = {http://dml.mathdoc.fr/item/1176350260}
}
Muller, Hans-Georg; Stadtmuller, Ulrich. Variable Bandwidth Kernel Estimators of Regression Curves. Ann. Statist., Tome 15 (1987) no. 1, pp.  182-201. http://gdmltest.u-ga.fr/item/1176350260/