Figure 7

Comparing the Jensen Shannon Divergence to the updated Bayes factor. Inference was performed based on data generated from the true model. Bayes factor updates pointing to the correct model were calculated given a new measurement which was based on the inferred posterior predictive distribution of the true model. Top row: Predictive distribution of updated Bayes factor in favor of the correct model with associated 67% credible intervals (dashed) for two models. Second row: Jensen Shannon divergence between the relevant Posterior Predictive Distributions. Bottom row: Relation between the updated Bayes factors and the JSD. Note that the relation is nearly monotonic indicating its usefulness as a predictor of experimental efficacy.