文摘
We define a new bandwidth-dependent kernel density estimator that improves existing convergence rates for the bias, and preserves that of the variation, when the error is measured in L1L1. No additional assumptions are imposed to the extant literature.