Thanh Mai Pham Ngoc (Laboratoire de mathématiques Université Paris Sud, Orsay)

Carte non disponible

Regression non paramétrique multivariée avec erreurs sur les variables

vendredi 14 octobre 2016, 9h30 - 10h30

Salle du conseil, espace Turing

In the multidimensional setting, we consider the errors-in-variables model. We aim at estimating the unknown multivariate regression function with errors in the covariates when the noise is ordinary smooth. We devise an adaptive estimator based on projection kernels on wavelet bases and a deconvolution operator. We propose an automatic and fully data driven procedure to select the wavelet level resolution. We obtain an oracle inequality and optimal rates of convergence over anisotropic Hölder classes. We illustrate our theoretical results with some simulations.