Title: Censored Poisson regression with missing censoring information
Authors: Bilel Bousselmi - IRMAR-INSA Rennes (France)
Jean-Francois Dupuy - INSA de Rennes (France) [presenting]
Abderrazek Karoui - Faculte des Sciences de Bizerte (Tunisia)
Abstract: Estimation in the Poisson regression model is considered when the count response is randomly right-censored and the censoring indicator can be missing at random. We investigate several estimation methods, such as multiple imputation and augmented-inverse-probability-weighted estimation. We derive the asymptotic properties of the resulting estimators (consistency, asymptotic normality, consistent variance estimation). A simulation study is conducted to evaluate and compare the proposed estimates.