A counting process approach to maximum likelihood estimation in frailty models

We study counting process models for event history data where the intensities depend on unobservable quantities (frailties). Examples include models for dependent failure times and regression models with unobservable covariates. Estimation in both parametric and nonor semi-parametric models is performed by maximum likelihood methods using the EM algorithm. Simulations and practical examples are presented.