Abstract:
|
Though the Cox proportional hazards (PH) model is popular in the analysis of survival data, it is difficult to use when the underlying survival function is of primary interest. In this case, Accelerated Failure-time models (AFT) can be used to estimate the survival function of interest; however, such models typically rely distributional assumptions that may be inappropriate outside the range of the data, or multiple AFT models may describe the data reasonably, but diverge in estimates in the tails of the survival function. Motivated by experiments on allergic response, where subjects are dosed with increasing amounts of a toxicant until a response is seen or the maximum dose is administered, A model averaged AFT model is presented that allows one to estimate the underlying survival function across multiple models. Simulation results as well as results from food allergy studies are presented using this new model.
|