B1357
Title: New approaches for bootstrap bandwidth selection in density estimation under dependence
Authors: Ines Barbeito - University of A Coruna (Spain) [presenting]
Ricardo Cao - University of Coruna (Spain)
Abstract: Smoothed versions of the Stationary Bootstrap and Moving Blocks Bootstrap are established for bandwidth selection in density estimation for dependent data. Exact expressions for the bootstrap version of the mean integrated squared error under dependence are obtained in both contexts. Those expressions are very useful, since the implementation of the bootstrap selector does not require Monte Carlo approximations. The empirical choice of the block length in these two situations is also analyzed. Finally, the good practical performance of both new bootstrap bandwidth selectors is shown in an extensive simulation study. The methods are illustrated by applying them to two real data sets.
Journal