Contributions to modeling duration data in the presence of censoring: application to the study of health insurance contract terminations.

Summary In this thesis, we are interested in duration models in the context of modeling the termination durations of health insurance contracts. Identified as early as the 17th century and the studies of Graunt J. (1662) on mortality, the bias induced by the censoring of duration data observed in this context must be corrected by the statistical models used. Through the problem of measuring the dependence between two successive durations, and the problem of predicting the duration of termination of an insurance contract, we study the theoretical and practical properties of different estimators based on a method of weighting the observations (the so-called IPCW method) aimed at correcting this bias. The application of these methods to the estimation of the customer value in insurance is also detailed.