ABSTRACT
The competing risks setting is integral to the modeling of multistate processes in which a transition out of some state can be to two or more other states. As in other settings involving inverse probability of censoring weights, these weights effectively create pseudo-risk sets that are representative of the risk sets that would be obtained had they not been chosen based on the censoring times. The pseudo-value method was proposed by Andersen et al. as a way to model and estimate covariate effects on a state occupancy probability at a particular time or set of times. Like the direct binomial regression approach, this framework is quite general in that it can be used for non-progressive multistate processes, but application to the competing risks setting is very natural. The chapter considers the analysis of times to failure and causes of failure for shunts inserted in 839 children with hydrocephalus.
