Split Sample Empirical Likelihood

TL;DR

This paper introduces a novel split sample empirical likelihood method that approximates the true likelihood function efficiently, maintaining asymptotic properties while reducing computational costs.

Contribution

The paper presents a new split sample approach to empirical likelihood that improves computational efficiency without sacrificing asymptotic accuracy.

Findings

Performs comparably to traditional empirical likelihood

Significantly reduces computational time

Maintains asymptotic properties of empirical likelihood

Abstract

We propose a new approach that combines multiple non-parametric likelihood-type components to build a data-driven approximation of the true likelihood function. Our approach is built on empirical likelihood, a non-parametric approximation of the likelihood function. We show the asymptotic behaviors of our approach are identical to those seen in empirical likelihood. We demonstrate that our method performs comparably to empirical likelihood while significantly decreasing computational time.