THE CHOICE OF KERNEL IN KERNEL DENSITY ESTIMATION

ABSTRACT

This study is centered on showing the nonparametric density estimation method for different choices of kernel, their efficiencies and the MISEs from different bandwidths. Three sets of data were analyzed with sample sizes of Forty, Sixty-four and One Hundred and Ten respectively. The density plots obtained for the different kernels generated different bandwidths for the three data sets. Choices of bandwidth considered were the subjective and Rule of thumb methods. The MISE table was generated and it depicts how good the smoothing parameter is. Efficiency of kernel was also considered which tells us how good the kernel is. The closer the efficiency to 1, the better the kernel.