Scholarlink

Double machine learning (DML) is a methodological framework that applies machine learning to reduce omitted variable bias in causal inference. Traditional regression models struggle to account for numerous confounders and to model non-linear relationships effectively. In contrast, DML can integrate multiple confounding variables while accurately capturing complex non-linear effects. It also helps identify heterogeneous causal effects and trace their dynamics over time. Furthermore, DML can be combined with causal mediation analysis, panel data modeling, and unstructured data analysis, demonstrating broad potential for quantitative social science research. This chapter outlines recent developments in DML, discusses its opportunities and challenges, and presents an empirical case on how parental academic expectations influence children’s academic performance.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Double Machine Learning for Causal Inference on High-Dimensional Data: A Flexible and Robust Approach to Causal Estimation

  • 章节

  • 出版模式：

    开放获取

  • 发表日期：

    2026年

查看全文
  • Yunsong Chen,
  • Zhuo Chen,
  • Wen Ma,
  • Guodong Ju

摘要

Double machine learning (DML) is a methodological framework that applies machine learning to reduce omitted variable bias in causal inference. Traditional regression models struggle to account for numerous confounders and to model non-linear relationships effectively. In contrast, DML can integrate multiple confounding variables while accurately capturing complex non-linear effects. It also helps identify heterogeneous causal effects and trace their dynamics over time. Furthermore, DML can be combined with causal mediation analysis, panel data modeling, and unstructured data analysis, demonstrating broad potential for quantitative social science research. This chapter outlines recent developments in DML, discusses its opportunities and challenges, and presents an empirical case on how parental academic expectations influence children’s academic performance.