The structure of a simulation heavily depends on the particular task. Often statistical simulation experiments are carried out with simplified conditions. For example, to judge a method, a univariate or multivariate normal distribution is used to simulate random numbers. The method is then applied to the simulated data. Such simulations often don't show the features of an estimation method, since the data structures are often much more complex in practice and it is very difficult to derive a real world behavior. So methods to simulate random numbers, as presented in Chapter 4, Simulation of Random Numbers may not be sufficient for complex simulation studies. For teaching, micro-simulation studies, remote execution tasks, and complex simulation studies, complex data must be simulated.
Usually, we speak about and carry out model-based simulation studies. In the model-based simulation world, first, data is drawn randomly by a super-population model...