Publication details
Kernel choosing with respect to the bandwidth in kernel density estimates
| Authors | |
|---|---|
| Year of publication | 2010 |
| Type | Conference abstract |
| MU Faculty or unit | |
| Citation | |
| Description | Kernel density estimates belong to the most popular nonparametric density estimates. It is a known fact, that these estimates depend on a bandwidth, which controls the smoothness of the estimate, and on a kernel, which plays a role of a weight function. Methods for finding the optimal bandwidth were presented e.g. in Duong & Hazelton (2005), Sain et al. (1994). We focus on the kernel function choice, especially on kernels with a bounded support. Wand and Jones (1995) suggested a criterion for the optimal kernel. Our aim is to study the optimality of the kernel with respect to the bandwidth choice. A simulation study brings comparison of the kernels. It shows up, that the cosine kernel performs better than the frequently used Epanechnikov kernel. |