Kernel density regression in the additive model: a B-spline approach

This paper investigates the estimation of the additive model. The nonparametric functions in the model are approximated through B-splines, and the kernel density regression method is employed to estimate the unknown parameters. Moreover, the convergence rate of the proposed approach is established. We conducted numerical experiments and real-world data analysis to validate the theoretical properties of our proposed method. Our numerical findings indicate that our approach offers superior estimation performance compared to several existing methods for the additive model, particularly in the presence of asymmetric, multimodal, or heavy-tailed error distributions.