|
Abstract:
|
A new class of frailty models is used to account for not only the correlation between failure times, but also the correlation between failure and censoring. These models have wide applications in the analyses for clustered survival data, recurrent event data and others. The proposed models allow multiple types of censoring. Censoring by some causes can be analyzed as informative, while censoring by other causes are assumed to be non-informative. To fit the model, the EM algorithm is used with Markov Chain Monte Carlo simulations in the E-steps. The EM algorithm in the setting of semi-parametric frailty models involves some complications, which will be illustrated and clarified in detail. Simulation studies show that the proposed models and the fitting method work well.
|
