The use of computers as an "experimentation environment" is becoming increasingly important in a wide area of applications. On the one hand, computers are used in experimental set-ups  such as a measuring, controlling, and data analysis devices, which are essential  to the accurate measurement and data handling. On the other  hand, we have the field where computers are used to perform some simulation of a "real-world"  phenomenon. One of the crucial components of that research field is the correct abstraction of a real-world  phenomenon to a conceptual model and the translation  into a computational  model that can be validated. This leads us to the notion of a computer experiment, where the model and the computer take the place of the "classical" experimental set-up where the simulation replaces the experiment.
In general, computer experiments are done for the following   purposes:

1. Simulations  of complex systems is when many parameters  must be determined before construction of a real system (e.g., car crash simulation).
2. Simulation of systems  in search  for phenomena not  directly  feasible  in laboratories.  For example, due to extremely  large or small time scales (e.g., stellar   clustering  evolutions) . It   may also not  be wise or possible to double the unemployment  rate  in reality  to determine the effect of employment on inflation,  while  a simulation  can be considered "harmless".
3. Simulation  of "theory  experiments" to increase  insight  in a theoretical  model.