Smoothed cross-validation

摘要: For bandwidth selection of a kernel density estimator, generalization the widely studied least squares cross-validation method is considered. The essential idea to do particular type “presmoothing” data. This seen be essentially same as using smoothed bootstrap estimate mean integrated squared error. Analysis reveals that rather large amount presmoothing yields excellent asymptotic performance. rate convergence optimum known best possible under wide range smoothness conditions. more appealing than other selectors with this property, because its motivation not heavily dependent on precise analysis, and form simple intuitive. Theory also given for choice presmoothing, used derive data-based choice.