|
Abstract:
|
Multistate statistical models describe multiple-event survival data that occur over time. They model stochastic processes that can progress through various states. The analysis focuses on modeling waiting times until events of interest occur. The theories are formalized within counting processes. While the theory is elegant, applications have been limited to Markov models. As an alternative, semi-Markov models have gained popularity recently. Among them, flowgraph models have demonstrated great flexibilities. Flowgraphs operate in moment generating function space, and allow virtually any waiting time distributions between transitions. Cumulative distribution function and probability density function are converted via saddlepoint approximations, The survivor and hazard are then computed. The method is illustrated using kidney transplant data.
|
