ABOUT CHOICE OF KERNEL IN NONPARAMETRIC ESTIMATION OF PROBABILITY DENSITY
The choice of weight function (kernel) by constructing the Parzen – Rosenblatt density estimator is discussed. The author specifies the conditions, under which the application of alternating weight functions (higher order kernels) improves essentially the quality of the probability density estimation. There are some references a reader can look through and find a weight function (a kernel) of required order with bounded support.
Keywords: nonparametric probability density estimator, alternating kernels (higher order kernels), an infinite order kernel (sinc-kernel), choice of kernel order.