Collaborative Research: Reducing Computation in Empirical Likelihood Methods

Georgia Tech Research Corporation, Atlanta GA

Investigators

Abstract

The study of empirical likelihood methods has attracted much attention in many areas of statistics. This proposal develops new procedures to overcome the computational burden in applying these methods to inference problems in different settings such as: estimating equations, stationary sequences, high-dimensional data, Pickands dependence functions in multivariate extreme-value distributions and copulas in risk management. A practical obstacle in using empirical likelihood methods is the computational issue due to the high dimensional data, the large size of nuisance parameters as well as dependence in the data. The investigators intend to develop new methodologies to overcome the computational problems and extend the scope of the potential applications. The new research results can be applied to economics, insurance, finance, risk management and other social sciences that require effective tools for exploring nonlinear dependence among multivariate series and to increase the complexity of the models.