Simulating simple and complex survival data

Simulation studies are conducted to assess novel and currently used methods in practice, to better assess and understand the frameworks under question. In survival analysis, we are interested in simulating both an event and a censoring distribution to better reflect clinical data. In this talk, I will describe how to simulate survival times from simple parametric distributions and then move to a more general framework, illustrating how to simulate from a general user-defined hazard function. This can incorporate any combination of a complex baseline hazard function with turning points, time-dependent effects, random effects, and nonlinear covariate effects. This is achieved through a two-stage algorithm incorporating numerical integration nested within root-finding techniques. The methods will be illustrated using the publicly available survsim package.