deux contributions à l'étude semi-paramétrique d'un modèle de régression

Abstract : In this thesis we are interested in two semiparametric regression models which allow to get rid with the so-called curse of dimensionality problem linked with the nonparametrical approach. The first part of this work deals with the study of the regression model called partialy-linear ; the goal is the identification of the regressor that appear nonparametricaly in the regression function and we also estimate the parameters defining the model. Then we define a linearity measure and we derive a test of the number of non-linear components based on this measure. The second part is devoted to the study of the so-called single index model and its aim is to estimate the axis using geometrical properties and to construct a test of the whole model consistent under the null hypothesis and powerfull under a sequence of local alternatives.
Document type :
Mathematics [math]. Université Rennes 1, 2003. French
Contributor : Céline Vial <>
Submitted on : Tuesday, March 8, 2005 - 11:02:24 PM
Last modification on : Tuesday, November 3, 2015 - 4:15:14 PM


  • HAL Id : tel-00008730, version 1



Céline Roget-Vial. deux contributions à l'étude semi-paramétrique d'un modèle de régression. Mathematics [math]. Université Rennes 1, 2003. French. <tel-00008730>




Consultation de
la notice


Téléchargement du document