Figure 4

Different steps of the proposed methodology. Different steps of the proposed methodology. Step 1: Different model topologies have to be formulated into a form with which simulations can be obtained. Step 2: Posterior distributions of parameters have to be sampled. In the panel, the joint posterior probability distributions of a few model parameters are depicted, with their marginals on the diagonal. Step 3: Use sample from the posterior parameter distribution to make predictions of quantities that can be measured (a sample of the Posterior Predictive Distribution). Step 4: Compute JSD between the predictive distribution of different models. Step 5: Perform the optimal experiment and compute the new Bayes factors. Optionally repeat until the hypotheses are sufficiently refined.