Effect of model complexity on marginal likelihood. Three different illustrative examples of integrated likelihoods. Left: Integrated likelihood under wide priors. The mismatch of the prior with respect to the high likelihood region results in low weights for the high likelihood region and therefore low model evidence. This situation is comparable to a case where the model contains too many parameters. A surplus of model parameters leads to a larger parameter space and therefore lower weights in high likelihood region. Middle: A good match between prior and likelihood. Right: A model that does not have sufficient freedom to describe the data very well.