Global Priors Ratio

It is possible to view the prior of a model, $P(M)$, as consisting of two parts: global and local. The local is the probability of the branch of computation that leading to $M$ and the global is an factor that can be computed after producing the model. As idicated by the name, we want to use the global part in computing the contribution of global features of the model to the prior. An example of the importance of such global features can be found in the prior used by the Familias program [Egeland, Pettera, Mevåg, StenersenEgeland et al.2000]. MCMCMSwill look into memory for the predicate global_priors_ratio/3. If this is present it will be called each time the acceptance probability is computed. The call is: $global\_priors\_ratio( +\ensuremath{{M_*}} , +\ensuremath{{M_i}} , -Ratio )$. The user-defined predicate is expected to return $Ratio = \frac{\ensuremath{P_G(\ensuremath{{M_*}} )} }{\ensuremath{P_G(\ensuremath{{M_i}} )} }$.