战略均衡系统因果效应的双鲁棒估计
机器学习
2026-04-03 v4
摘要
我们引入了战略双鲁棒（SDR）估计器，这是一种新型框架，将战略均衡建模与双鲁棒估计集成，用于战略环境中的因果推断。SDR 解决了源于战略代理人行为的内在性处理分配问题，同时保持双鲁棒性并纳入战略考虑。理论分析确认 SDR 在战略无偏 confounding 下的一致性和渐近正态性。实证评估表明，SDR 在不同战略强度下优于基线方法，实现 7.6%-29.3% 的偏差降低，并在代理人人口规模上保持鲁棒的可扩展性。该框架为在代理人作出战略反应时进行可靠因果推断提供了原则方法。
引用
@article{arxiv.2510.15555,
title = {Doubly Robust Estimation of Causal Effects in Strategic Equilibrium Systems},
author = {Sibo Xiao},
journal= {arXiv preprint arXiv:2510.15555},
year = {2026}
}
备注
In systems with causal effects, a large majority of individuals are mistakenly classified as using a certain strategy by the strategic equilibrium solver, resulting in the introduction of this feature as an independent variable in causal inference without specificity. This method may have an inherent error