Partial BSR simulation and reestimation

[Víctor de Buen Remiro]

Most of applications of BSR models in inference and forecast tasks, and also reestimation of models to update small changes as a few new data or variables, may be treated as simulation process that could be defined as simplified versions of original and previously simulated model.

If we have an stored copy of definition and simulation results of a model and we have a simplified method to modify it in some predefined senses, then we can use BSR in a faster way to reestimate, forecast or inference tasks.

In order to reduce computational cost, user should be able to fix some parameters to specific values or skip the simulation of other ones simply by reading them from stored MCMC.