Theoretical Background

Abstract

In biometric studies the Cox’s proportional hazards model (Cox, 1972) has been established as the most popular regression model for analysing survival data. This kind of data is characterized by two parts. On the one hand by failure, which describes the occurrence of an event such as death or a specific disease, and on the other hand by failure time, which is the time period until the event occurs. A special property of survival data is the presence of censoring. A subject is called as censored, when it drops out early without having an event or is event free by the end of the study. This type of censorship is also called right censoring. Here, only this type of censorship is considered. For a detailed overview of the different censoring mechanisms see e. g. Kalbfleisch and Prentice (2002).