| DATE: | Thursday, March 21, 2002 |
| TIME: | 4:30-5:30 PM |
| PLACE: | LN 2205 |
| SPEAKER: | W. Wefelmeyer, University of Siegen, Germany |
| TITLE: | Plug-in Estimators for Regression Models and Time Series |
Smooth functionals of densities or regression functions (for example a distribution function or a regression function average) can be estimated by plugging an appropriate kernel estimator into the functional. This typically improves the rate of convergence; in certain cases one even obtains an efficient estimator. We consider in particular efficient estimation of the error variance and the error distribution in nonparametric regression. Rather surprisingly, in regression-type models the plug-in method can also be used to construct estimators for densities and conditional expectations that converge faster than the usual estimators. We describe such results for moving average and autoregressive time series. This is joint work with Uschi Mueller and Anton Schick.