Impact of Bayesian Approach on Kernel Conditional Density Estimator

Produkt bewerten

Nour El-Hayet Ladaouri

▪Mouloud Cherfaoui

Verlag:

Lambert Academic Publishing

Unsere-Artikel-Nr.: P36905396

EAN: 9786630222739

Erhältlich:

Nicht auf Lager

Zustellung: Mi, 26.08.2026

Versand: Kostenlos

-11.6 %

CHF 86.–

CHF 76.–

Beschreibung

This book focuses on the kernel estimation of the conditional density function. It is particularly interested in the influence of the choice of smoothing parameter on the performance of this estimator. The results obtained through this work show that: the problems associated with the unbiased cross-validation method in the conditional density estimation remains the same as in the univariate density. Moreover, to have a more performance estimator in the sense of the ISE it is preferable to impose the hypothesis of non-equality of the two smoothing parameters, however, if the goal is to minimize the calculation time, it is preferable to impose equality between them. Furthermore, the conditional density estimator, when the explanatory variable is functional using a symmetric kernel can give reasonable results regardless of the norm used, indeed, when we changed the norm we obtained estimators with approximately the same average ISE. The results also proved that, when the sample size is small or medium, the Bayesian alternative is more efficient than the classical method, but this alternative requires significantly more computation time to implement compared to classical methods.