Probability and Statistics Seminar: George Aivaliotis (Seminar)

Non-linear discrete-time survival models are among the flexible survival models
in which the proportional hazard assumption is not imposed. We propose several
discrimination measures to assess the models' predictive performance, including
time-dependent Uno's C-index and pair calibration, which the former outperforms
the standard Antolini's time-dependent concordance. We investigate the proposed
discrimination measures using real and simulated data.