Beyond Monte Carlo
Jochen Voß · 2013
This chapter presents two methods which can be used instead of Monte Carlo methods if either no statistical model is available or if the available models are too complicated to easily apply Monte Carlo methods. The first of these two methods, called Approximate Bayesian Computation (ABC), is tailored towards the case where a computer model of the studied system is available, which can be used instead of a mathematical model. ABC is a computational technique to obtain approximate parameter estimates in a Bayesian setting. The ABC approach, described in the chapter, is less efficient than MCMC methods are, but ABC is easier to implement than MCMC and the method requires very little theoretical knowledge about the underlying model. The second method, Bootstrap sampling method, is used in cases where only a set of observations but no model at all is available.