There is an increasing amount of literature focused on Bayesian computationalmethods to address problems with intractable likelihood. One approach is a setof algorithms known as Approximate Bayesian Computational (ABC) methods. One ofthe problems of these algorithms is that the performance depends on the tuningof some parameters, such as the summary statistics, distance and tolerancelevel. To bypass this problem, Mengersen, Pudlo and Robert (2013) introduced analternative method based on empirical likelihood, which can be easilyimplemented when a set of constraints, related to the moments of thedistribution, is known. However, the choice of the constraints is sometimeschallenging. To overcome this problem, we propose an alternative method basedon a bootstrap likelihood approach. The method is easy to implement and in somecases it is faster than the other approaches. The performance of the algorithmis illustrated with examples in Population Genetics, Time Series and StochasticDifferential Equations. Finally, we test the method on a real dataset.
展开▼
