ABSTRACT
Classical competing risks models are discussed, as well as their modern interpretations. The basic problemis that we want to consider more than one type of event in situations where exactly one will occur. Forinstance, consider death in different causes. The first problem is: How can the hazards be nonparametricallyestimated? It turns out that this is quite simple. The trick is to take one cause at at time and estimateits intensity as if the rest of the causes (events) are censorings. The real problem starts when we want tohave these intensities turned into probabilities, and what kind of probabilities that are possible to estimateis discussed.
