方法对比

	双重机器学习	异质性处理效应（CATE / 元学习器）	随机森林
领域	因果推断	因果推断	机器学习
方法族	Machine learning	Regression model	Machine learning
起源年份	2018	2018	2001
提出者	Victor Chernozhukov et al.	Wager & Athey (causal forest); Künzel et al. (meta-learners)	Breiman, L.
类型	Semiparametric causal estimation	Causal machine-learning framework	Ensemble (bagging of decision trees)
开创性文献	Chernozhukov, V., Chetverikov, D., Demirer, M., Duflo, E., Hansen, C., Newey, W., & Robins, J. (2018). Double/debiased machine learning for treatment and structural parameters. The Econometrics Journal, 21(1), C1–C68.	Wager, S. & Athey, S. (2018). Estimation and Inference of Heterogeneous Treatment Effects using Random Forests. Journal of the American Statistical Association.	Breiman, L. (2001). Random Forests. Machine Learning, 45, 5–32.
别名	Debiased Machine Learning, Neyman Orthogonal Score Estimation, Partialing-Out Lasso, Çift Makine Öğrenmesi	conditional average treatment effect, CATE, meta-learners, causal forest	Rastgele Orman (Random Forest), rastgele orman, random decision forest, bagged tree ensemble
相关	3	5	4
摘要	Double/Debiased Machine Learning (DML), introduced by Chernozhukov et al. (2018), is a semiparametric framework for estimating causal or structural parameters in the presence of high-dimensional controls. It uses flexible machine learning methods to model nuisance functions—the conditional expectations of the outcome and the treatment given covariates—and then constructs a debiased estimator of the target parameter that achieves root-n consistency and valid inference despite the regularization bias inherent in high-dimensional settings.	Heterogeneous Treatment Effects is a machine-learning framework that estimates how a treatment effect varies across individuals — the conditional average treatment effect (CATE). It bundles meta-learner strategies such as the T-Learner, S-Learner, X-Learner and R-Learner alongside the causal forest of Wager and Athey (2018) and Künzel et al. (2019).	Random Forest is an ensemble learning method, introduced by Leo Breiman in 2001, that grows many decision trees on bootstrap samples of the data and combines their votes to produce strong classification and regression. By pooling many slightly different trees, it produces more accurate and more stable predictions than any single tree.