Data Fusion Using Robust Empirical Likelihood Inference

TL;DR

This paper introduces a robust semi-parametric empirical likelihood method for data fusion from multiple samples, improving inference accuracy without needing detailed distributional assumptions.

Contribution

It presents a novel empirical likelihood approach that enhances data integration and inference efficiency in multi-sample settings without requiring known distribution forms.

Findings

Outperforms classical likelihood in simulations

Provides more accurate confidence intervals

Offers efficient inference under structural parameter relationships

Abstract

The authors propose a robust semi-parametric empirical likelihood method to integrate all available information from multiple samples with a common center of measurements. Two different sets of estimating equations are used to improve the classical likelihood inference on the measurement center. The proposed method does not require the knowledge of the functional forms of the probability density functions of related populations. The advantages of the proposed method were demonstrated through the extensive simulation studies by comparing mean squared error, coverage probabilities and average length of confidence intervals with those from the classical likelihood method. Simulation results suggest that our approach provides more informative and efficient inference than the conventional maximum likelihood estimator when certain structural relationships exist among the parameters for these relevant samples.