[论文] When Prediction Error Is Not Enough: Evaluating Nuisance-Function Pred...

论文概要

研究领域: ML
作者: Cong Cao
发布时间: 2026-09-03
arXiv: 2509.00010

中文摘要

预测误差被广泛用于评估因果推断中的干扰函数估计器，但它与因果估计器性能的关系可能因性能度量而异。我们使用蒙特卡洛模拟在部分线性模型中研究了这个问题。我们比较了普通最小二乘法（OLS）、广义可加模型（GAM）、XGBoost和XGBoost双机器学习（DML-XGBoost），评估干扰函数预测误差、偏差、RMSE和95%置信区间覆盖率。我们还检查了一个基于暴露和结果干扰函数估计误差绝对叉积的简单联合误差度量。跨模拟设置，XGBoost在非oracle方法中具有最低的RMSE，而DML-XGBoost通常提供更好的置信区间覆盖率。预测误差在方法和设置间不能一致地跟踪因果偏差，具有最佳点估计性能的方法不一定具有最佳的置信区间覆盖率。联合误差度量仅与因果偏差弱相关，并未提供有用的因果性能独立度量。这些结果表明，预测误差对评估干扰函数估计是有用的，但不应被视为结果因果估计器质量的直接度量。

原文摘要

Prediction error is widely used to evaluate nuisance-function estimators in causal inference, but its relationship with causal estimator performance may differ across performance measures. We studied this question in a partially linear model using Monte Carlo simulations. We compared ordinary least squares (OLS), generalized additive models (GAMs), XGBoost, and Double Machine Learning with XGBoost (DML-XGBoost), evaluating nuisance-function prediction error, bias, RMSE, and 95% confidence interval coverage. We also examined a simple joint-error measure based on the absolute cross-product of estimation errors from the exposure and outcome nuisance functions. Across the simulated settings, XGBoost had the lowest RMSE among the non-oracle methods, while DML-XGBoost generally provided better c...

自动采集于 2026-09-03

#论文 #arXiv #ML #小凯