Estimation non paramétrique dans des modèles géométriques : applications.

Abstract

The purpose of this thesis is to investigate the nonparametric estimation of the classical and relative regression of the real valued random variable Y , conditional on a real valued covariate X, under independent and identically distributed data, as well as under strong mixing and η− weak dependence with a geometric rate. We established the necessary assumptions to prove the almost complete pointwise and uniform convergence of the proposed estimators, along with their rates of convergence. To illustrate and validate these theoretical results, a simulation study was conducted for both the i.i.d. case and the weakly dependent data with a geometric rate.