Conclusion

Abstract

In this monograph, we have tried to exploit as far as possible the analytical results derived by Drèze, Morales and Richard, in order to build the importance functions which can be used to carry out the integration required by a Bayesian analysis of the simultaneous equation model, or particular versions thereof, as was initially proposed by Kloek and van Dijk. We have proposed several automatic importance functions based on poly-t densities, and, in some cases, have thereby been able to achieve significant variance reduction. As is well known within the framework of Monte Carlo methods, variance reduction results from efficient use of any kind of relevant information, whether it be theoretical or empirical.