Extent to which least-squares cross-validation minimises integrated square error in nonparametric density estimation (Q1073495)

Let \(h_ 0,\hat h_ 0\) and \(\hat h_ c\) be the windows which minimise mean integrated square error, integrated square error and the least- squares cross-validatory criterion, respectively, for kernel density estimates. It is argued that \(\hat h_ 0\), not \(h_ 0\), should be the benchmark for comparing different data-driven approaches to the determination of window size. Asymptotic properties of \(h_ 0-\hat h_ 0\) and \(\hat h_ c-\hat h_ 0\), and of differences between integrated square errors evaluated at these windows, are derived. It is shown that in comparison to the benchmark \(\hat h_ 0\), the observable window \(\hat h_ c\) performs as well as the so-called ''optimal'' but unattainable window \(h_ 0\), to both first and second order.