A Short Note on Finite Sample Analysis on Double/Debiased Machine Learning

Yonghan Jung

Keywords: Double/Debiased Machine Learning, Finite Sample Error Analysis

TL;DR: We provided a short discussion on the finite sample error analysis on double/debiased machine learning inference framework.

Abstract: This note provides learning guarantees for sample-splitting-based estimators, which include double/debiased machine learning (DML) (Chernozhukov et al., 2018) estimators. We prove consistency and Gaussian approximation of estimators using finite-sample arguments, extending the general asymptotic theory. Our work extends previous research (Chernozhukov et al., 2023; Quintas-Martinez, 2022) that studied learning guarantees for the expected linear functional in general sample-splitting-based estimators.