No label defined (Q4236282)

This CWI tract is the Ph.D. dissertation of the author. It is a well known fact that the proof of consistency of estimators is closely related to the uniform convergence of the sequence of criteria functions used to define the corresponding estimator. Therefore the theory of empirical processes is an adapted tool to study the problem of consistency. After giving a short introduction to empirical processes, the case of finite dimension parameter spaces is treated. But the main goal of the book concerns the nonparametric regression model both for random and nonrandom design. S. van de Geer proved the consistency of the least squares estimator under metric entropy conditions on the set of regression functions under consideration. These conditions are shown to be necessary in chapter 4 of the book. Furthermore, under moment conditions, a lower bound for the rate of convergence has been established. In the special case of Gaussian errors this rate is shown to be optimal. The book concludes with limit theorems for both the empirical norm and the estimator. The book provides both an introduction to basic ideas of empirical processes, an overview on recent results in the field of nonparametric regression and a large number of new results obtained by the author.