Title: Parametric and semiparametric modeling of doubly truncated lifetimes under time-restricted data collection schemes
Authors: Achim Doerre - University of Rostock (Germany) [presenting]
Abstract: Double truncation occurs when individuals are observed if and only if their variable of interest is within two, possibly random, truncation bounds. In survival analysis, double truncation has been an ongoing topic due to multiple applications in areas including medicine, engineering and economics. The quasi-independence of the truncation variables and the variable of interest, an essential and critical assumption of many methods for truncation, is studied within a multidimensional Poisson process modeling framework. We primarily focus on a type of double truncation which is induced by time-restricted data collection, for which we argue that conventional nonparametric methods often have limited practical capacity. Additional considerations concerning asymptotics and the available parametric and semi-parametric modeling alternatives further motivate the need for a suitable goodness-of-fit assessment. We investigate the latter via Bayesian methods. Finally, we apply the proposed methods to a dataset of German companies whose age at insolvency is recorded conditional on their insolvency event.