Analysis of Clinical Trials Using SAS

1.4.2 Model-Based Tests

As in the case of categorical outcomes considered in Section 1.3, randomization-based tests for time-to-event data generally apply only to rather simple statistical problems, e.g., testing for equality of survival functions. Therefore, if we are interested in investigating effects of mixed (continuous and categorical) covariates on survival distributions or studying the prognostic ability of multiple factors, we need to rely on more flexible methods based on regression models for time-to-event data. Time-to-event outcome variables are analyzed using parametric models that explicitly specify the underlying survival distribution or using semi-parametric models such as proportional hazards models. Parametric regression ...