Model hyperparameters controlling the scaling and bias of the latent function were optimized by maximising the model evidence, a procedure also referred to as type-II maximum likelihood (see [Marquand et al., 2010] for full details).
The phase diagram of the pure model evidences a rich polymorphism, with isotropic, columnar and crystalline phases at low pressure, and the appearance of nematic phase at higher pressure.
Likewise, the MAM model evidences a smaller thalamus and has now replicated data on MDT size reductions [45] (Figure 1D).
Here we provide in an animal model evidence for a therapeutic effect of environmental enrichment and exercise in the treatment of PTSD.
Hence, I report sensitivity in terms of a log Bayes factor (LBF) – the difference in negative log model evidence between a given model and a reference model.
CreatS correlated also negatively with CdU, and both models evidenced an interaction between PbB and HgU increasing CreatS.
Here, I use the Log Bayes Factor (LBF) as a measure of relative model evidence, where a smaller LBF indicates better model evidence, and in the present context this means higher predictive validity.
Moreover, the optimal full hierarchical Bayesian learning model showed higher model evidence than a Rescorla Wagner learning model or a model which assumes that the subjects knew the true underlying probabilities.
As model parameters are integrated out in this procedure, the model evidence includes a complexity term that penalizes models with more parameters, e.g., more columns in the design matrix in a general linear model.
Model fitting is achieved by adjusting model parameters to maximize the free-energy estimate (F) of the log model evidence for a given data set (Friston et al. 2003), adjusting for model complexity (in terms of both the number of parameters and dependencies among parameters).
