Non-Parametric Inference of Window Observation Recurrent Event Data with Multiple Causes of Failure

Recurrent event data with multiple causes of failure are common in survival studies. There are situations where the recurrence history of some systems is available in disconnected ob4 servation windows with gaps in between. Such data are called window-observation recurrence data, and no events can be observed during these gaps. This article discusses the non6 parametric estimation of cause specific mean function for window-observation recurrence data with multiple failure modes. A test statistic is proposed to compare the effect of dif8 ferent causes. Simulation studies are performed to evaluate the performance of the proposed methods. An automobile warranty database is analyzed to illustrate the suggested techniques.