Title: A generic framework for recurrent event data based on virtual age models and implemented in the R package VAM
Authors: Laurent Doyen - Univ Grenoble Alpes (France) [presenting]
Remy Drouilhet - Univ Grenoble Alpes (France)
Abstract: Virtual age models are useful to analyze recurrent events arising in epidemiology (e.g. relapse and treatment times of a disease), industry (e.g. failure and maintenance times of a system), etc. The model consists of a composition of a baseline hazard rate function, characterizing the first time to event distribution, and an effective age function, that allows to take into account events effects. We have developed, and implemented in the R package VAM, a general framework that allows to take into account simultaneously several types of events, having different effects and corresponding to different types of treatments or maintenances. We can also consider planned event times for which the next possible arising time is fixed function of the previous events times and types. The proposed framework is generic in the sense that we propose an iterative way for computing the different characteristics of the model that does not a priori depends on the number of different event types and of their effects. In order to preserve this adaptability in the software implementation, the usage of VAM is based on a formula which specify the characteristics of the data set to analyze and the model used for that. Methods are proposed for events times simulation, maximum likelihood estimation, reliability or health indicators computation. An application to a real data set issued from off-road engines of mining trucks will be presented.