Figure 2.
Conceptual diagram of the Monte Carlo method for model choice. First, parameters for both models are estimated from the original data. Then, n simulated datasets are created from each model at these parameters, and on each dataset, the parameters for both models are re-estimated and the likelihood ratio statistic is computed. The collection of likelihood ratio statistics generates the corresponding distribution. This involves a process of 4n fits by maximum likelihood, instead of only 2 fits required for information criteria.