This website requires JavaScript.

DeepMed: Semiparametric Causal Mediation Analysis with Debiased Deep Learning

Siqi XuLin LiuZhonghua Liu
Oct 2022
摘要
Causal mediation analysis can unpack the black box of causality and istherefore a powerful tool for disentangling causal pathways in biomedical andsocial sciences, and also for evaluating machine learning fairness. To reducebias for estimating Natural Direct and Indirect Effects in mediation analysis,we propose a new method called DeepMed that uses deep neural networks (DNNs) tocross-fit the infinite-dimensional nuisance functions in the efficientinfluence functions. We obtain novel theoretical results that our DeepMedmethod (1) can achieve semiparametric efficiency bound without imposingsparsity constraints on the DNN architecture and (2) can adapt to certain lowdimensional structures of the nuisance functions, significantly advancing theexisting literature on DNN-based semiparametric causal inference. Extensivesynthetic experiments are conducted to support our findings and also expose thegap between theory and practice. As a proof of concept, we apply DeepMed toanalyze two real datasets on machine learning fairness and reach conclusionsconsistent with previous findings.
展开全部
图表提取

暂无人提供速读十问回答

论文十问由沈向洋博士提出,鼓励大家带着这十个问题去阅读论文,用有用的信息构建认知模型。写出自己的十问回答,还有机会在当前页面展示哦。

Q1论文试图解决什么问题?
Q2这是否是一个新的问题?
Q3这篇文章要验证一个什么科学假设?
0
被引用
笔记
问答