A Monte Carlo study into time-aggregation in continuous and discrete- time hazard models

The purpose of this study is to investigate the effect of time- aggregation in discrete and continuous-time hazard models. A Monte Carlo study is conducted in which data are generated according to underlying continuous and discrete-time processes, which data are aggregated into daily, weekly and monthly intervals. Under each of these conditions 400 datasets are generated, that vary in the parameters of the baseline hazard. These datasets are analyzed with flexible continuous-time and discrete-time proportional hazard models. The estimates of the structural parameter and of the baseline hazard, as well as the baseline hazard predicted by those estimates, seem robust to the form of the distribution of the data generation process when the time-aggregation window is small. Both estimates of continuous-time models and of discrete-time models suffer from time-aggregation, but the estimates of the discrete-time model are more sensitive to aggregation.