Asymptotics of the pseudo-observation method

Morten Overgaard, Section for Biostatistics, Aarhus University

Using jack-knife pseudo-observations as an outcome variable in a regression analysis has become known as the pseudo-observation method. This presentation is about when and why we should expect this approach to work. When looking at pseudo-observations from the Kaplan—Meier estimator or similar in a survival setting, the ‘when’ part involves a strict assumption on the censoring mechanism. The ‘why’ part can be answered by considering the estimator as a very well-behaved function of the empirical distribution.