Figure 2 : Random-Forest Approximate Bayesian Computation model-choice cross-validation. Heat map of the out-of-bag cross-validation results considering each 10,000 simulations per each nine competing models (Figure 1 , Table 1 ) in turn as pseudo-observed target for RF-ABC model-choice. Prior probability of correctly choosing a given scenario is 11%. Out-of-bag prior error rate is 32.41%. RF-ABC model-choice performed using 1,000 decision trees and 24 summary-statistics (see Materials and Methods ).