Skip to Main content Skip to Navigation

Vraisemblance empirique généralisée
et estimation semi-paramétrique

Abstract : Empirical likelihood is an estimation method inspired by the classical likelihood method, but without assuming any parametric model for the distribution of the data. The empirical likelihood method can be described as the maximization of the likelihood of a discrete distribution supported by the data. It can be used to build confidence regions, as long as the parameter of interest is defined by some moment constraints.
In this thesis, we will generalize the empirical likelihood method to a wide family of empirical discrepancy methods. We give in particular non asymptotic results for some well-chosen discrepancies. We will also propose an extension of empirical likelihood to Markov chains. Those theoretical results will be used in two. The first one proposes to evaluate some risk index for the exposition to methyl-mercury via sea products consumption, by taking into account several data sources. The second one evaluates the effect of social norm on obesity.
Document type :
Complete list of metadata

Cited literature [192 references]  Display  Hide  Download
Contributor : Hugo Harari-Kermadec Connect in order to contact the contributor
Submitted on : Tuesday, December 19, 2006 - 6:53:34 PM
Last modification on : Thursday, October 21, 2021 - 3:16:22 PM
Long-term archiving on: : Thursday, September 20, 2012 - 4:20:18 PM


  • HAL Id : tel-00121233, version 1
  • PRODINRA : 252087


Hugo Harari-Kermadec. Vraisemblance empirique généralisée
et estimation semi-paramétrique. Mathématiques [math]. ENSAE ParisTech, 2006. Français. ⟨tel-00121233⟩



Record views


Files downloads