Skip to Main content Skip to Navigation

Modèles statistiques pour des données de survie corrélées

Abstract : Most of the statistical models and methods for failure time data were implicitely developped under the assumption that the observations from subjetcs are statistically independant of each other. While sensible in many applications, this assumption is obviously violated in other situations which are not as uncommon as originally thought. For example, in veterinary science, such correlation between data occurs, specially when individuals recording single outcomes are grouped into clusters.
We study the two broad classes of models for correlated survival data: frailty (or conditional) models and marginal models. We propose a wide comparison of these two approaches; this comparison is realized through veterinary data set and simulations.
Our goal is to assess the sensitivity of such models, and more particulary to appreciate their dependance on several parameters (for example the clusters size).
Document type :
Complete list of metadata

Cited literature [38 references]  Display  Hide  Download
Contributor : Tristan Lorino <>
Submitted on : Monday, November 3, 2003 - 9:26:15 AM
Last modification on : Friday, October 23, 2020 - 4:52:18 PM
Long-term archiving on: : Friday, April 2, 2010 - 7:44:47 PM


  • HAL Id : tel-00003672, version 1



Tristan Lorino. Modèles statistiques pour des données de survie corrélées. Sciences du Vivant [q-bio]. Institut national agronomique paris-grignon - INA P-G, 2002. Français. ⟨tel-00003672⟩



Record views


Files downloads