2014/07/08_Anouar El Ghouch (Université Catholique de Louvain)

Prediction and variable selection in AFT cure models with interval censored data

2014/07/08 – 12:00 h | Anouar El Ghouch, Université Catholique de Louvain

Abstract

In this talk, we study an accelerated failure time (AFT) regression model when the data are subject to a general interval censoring scheme and in the presence of immune or cured individuals. We utilize an extended generalized gamma distribution for the error term of the AFT model and a logistic regression for the cure proportion (i.e. the incidence part). We investigate some issues concerning this model, including estimation, variable selection, model simplification and prediction. We then apply this method to an Alzheimer disease database, which consists in 241 at-risk patients followed-up between 1998 and 2008 with regular checks for the appearance of mild cognitive impairment.