CMStatistics 2016: Start Registration
View Submission - CFE
A0496
Title: Maximum likelihood method for bandwidth detection of kernel conditional density estimation Authors:  Ivanka Horova - Masaryk University (Czech Republic)
Katerina Konecna - Masaryk University (Czech Republic) [presenting]
Abstract: Kernel smoothing technique is a suitable tool for estimation of conditional density estimation. Kernel estimations depend on a kernel which plays a role of weight function and smoothing parameters which control a smoothness of the estimation. The problem of a choice of smoothing parameters, i.e., how much to smooth, is the most significant. The aim is to compare the well-known cross-validation method and maximum likelihood method. The performance of these methods is compared by simulation studies. Application to real data is also included.