🤖 AI Summary
This study addresses the limitations of traditional fixed-effects models, which rely on strong assumptions of linear additivity and independence, thereby struggling to accommodate group heterogeneity and within-group dependence and leading to biased cross-group comparisons. To overcome these issues, the authors propose a Graph Neural Network–based Generalized Mundlak Estimator (GME-GNN) that dispenses with conventional intercept terms and instead incorporates group-level balancing statistics to control for between-group confounding. By leveraging the message-passing mechanism of graph neural networks, the method adaptively learns nonlinear representations to flexibly capture intra-group interaction structures. The estimator is theoretically shown to possess double robustness and asymptotic normality. Both simulation experiments and empirical analyses demonstrate its superior performance over existing approaches in bias reduction and cross-group inference.
📝 Abstract
This paper proposes a generalized Mundlak estimator based on graph neural networks (GME-GNN). The estimator is designed to mitigate bias arising from group-level heterogeneity and to accommodate within-group dependence among individuals. Traditional fixed-effects models handle group heterogeneity via group-specific intercepts, but require overly strict linear additivity and intra-group independence assumptions, and are confined to within-group comparisons. Rather than relying on intercepts, GME-GNN uses aggregated group-level balancing statistics to fully control between-group confounding, enabling valid cross-group comparisons and relaxing linearity constraints. It further employs graph neural network message-passing to adaptively learn nonlinear representations and capture intra-group interaction effects. Theoretical analysis shows that the estimator satisfies double robustness and is asymptotically normal. Simulation and empirical studies confirm its performance.