Bayesian Empirical Likelihood

• Activity Number: 67
• Type: Topic Contributed
• Date/Time: Sunday, July 31, 2011 : 4:00 PM to 5:50 PM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Sanjay Chaudhuri*+
• Companies: National University of Singapore
• Keywords: Empirical likelihood ; Bayesian Statistics ; semi-parametric methods

Abstract

There has been a lot of interest about empirical likelihood in recent times. As a statistical methodology it has several advantages. First of all, it is a semi-parametric method and does not require any distributional assumption. Still, parametric models can be specified as constraints through estimating equations and the parameters can be estimated by constrained maximization of the likelihood. It is known that the parameters estimate are highly efficient and have usual asymptotic properties. Furthermore, empirical likelihood based estimates are easy to compute which can be very advantageous in many constrained problems. In this talk we illustrate the use of empirical likelihood as an alternative to traditional parametric likelihoods in Bayesian methodologies. In Bayesian framework, empirical likelihood has several advantages. In many applications it can handle both discrete and continuous data in an unified manner. Furthermore, restrictions to certain classes of estimators, often assumed for convenience, can also be dispensed with. We discuss the formulation and computational methodologies for the proposed estimator. Performance of the estimator is illustrated through examples.