BFS 2002 |
|
Contributed Talk |
Eldar Nigmatullin
In the presence of model uncertainty Bayesian model averaging (BMA) is a useful tool for accounting for model risk. The paper considers the practical implementation of BMA when models are specified only via moment conditions and no further assumptions are made about parametric specification of the probabilistic structure underlying the data sampling process. The paper proposes calculation of Bayes factors using empirical likelihood. Application of the proposed technique to portfolio choice problem is considered.