Multivariate density estimation with general flat-top kernels of infinite order (Q1283845)

The paper addresses the problem of nonparametric estimation of multivariate density functions. In particular, a general class of estimators with favourable asymptotic performance (bias, variance, and rate of convergence) is proposed by the authors. These proposed estimators are characterised by the flatness near the origin of the Fourier transform of the kernel and are actually shown to be exactly \(\sqrt N\)-consistent provided the density is sufficiently smooth.