A General Class of Additive Semiparametric Models for Recurrent Event Data
Recurrent event data is a special case of multivariate lifetime data that is present in a large variety of studies from numerous disciplines. Due to its pervasiveness, it is essential that appropriate models and inference procedures exist for its analysis. We propose a general class of additive semiparametric models for examining recurrent event data that uses an effective age process to take into account the impact of interventions applied to units after an event occurrence. The effect of covariates is additive instead of the common multiplicative assumption. We derive estimators of the regression parameter, baseline cumulative hazard function, and baseline survival function. We also establish the asymptotic properties of the estimators using tools from empirical process theory. Simulation studies indicate that the asymptotic properties of the regression parameter closely approximate its finite sample properties. The analysis of a real data set consisting of indolent lymphoma recurrence times provides a practical illustration of the class of models and is used to examine the impact of the effective age process. The importance of the effective age process is also demonstrated via the modeling of a data set of failure times for the hydraulic subsystems of load-haul-dump machines used in mining.
R. Stocker and A. Adekpedjou, "A General Class of Additive Semiparametric Models for Recurrent Event Data," Journal of Statistical Planning and Inference, vol. 205, pp. 231 - 244, Elsevier B.V., Mar 2020.
The definitive version is available at https://doi.org/10.1016/j.jspi.2019.07.006
Mathematics and Statistics
Keywords and Phrases
Additive model; Counting processes; Effective age process; Empirical process theory; Recurrent events
International Standard Serial Number (ISSN)
Article - Journal
© 2020 Elsevier B.V., All rights reserved.
01 Mar 2020