Kernel Density Estimation for Demonstration Purposes

Demonstration code showing how (univariate) kernel density estimates are computed, at least conceptually, and allowing users to experiment with different kernels, should they so wish. The method used follows directly the definition, but gains efficiency by replacing the observations by frequencies in a very fine grid covering the sample range. A canonical reference is B. W. Silverman, (1998) . NOTE: the density function in the stats package uses a more sophisticated method based on the fast Fourier transform and that function should be used if computational efficiency is a prime consideration.