Assumption-lean falsification tests of rate double-robustness of double-machine-learning estimators

DOI10.1016/J.JECONOM.2023.105500MaRDI QIDQ6199660zbMATH OpenFDO

Authors Lin Liu, Rajarshi Mukherjee, James Robins

Publication date 21 March 2024

Published in Journal of Econometrics (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/2306.10590

zbMATH Keywords

causal inferencedoubly robust functionalseconometricshigher-order \(U\)-statisticshigher-order influence functionsmachine learning

Mathematics Subject Classification ID

Statistics (62-XX) Game theory, economics, finance, and other social and behavioral sciences (91-XX)

Abstract: In this article we develop a feasible version of the assumption-lean tests in Liu et al. 20 that can falsify an analyst's justification for the validity of a reported nominal $(1-alpha)$ Wald confidence interval (CI) centered at a double machine learning (DML) estimator for any member of the class of doubly robust (DR) functionals studied by Rotnitzky et al. 21. The class of DR functionals is broad and of central importance in economics and biostatistics. It strictly includes both (i) the class of mean-square continuous functionals that can be written as an expectation of an affine functional of a conditional expectation studied by Chernozhukov et al. 22 and the class of functionals studied by Robins et al. 08. The present state-of-the-art estimators for DR functionals $psi$ are DML estimators $hat{psi}_1$. The bias of $hat{psi}_1$ depends on the product of the rates at which two nuisance functions $b$ and $p$ are estimated. Most commonly an analyst justifies the validity of her Wald CIs by proving that, under her complexity-reducing assumptions, the Cauchy-Schwarz (CS) upper bound for the bias of $hat{psi}_1$ is $o(n^{-1/2})$. Thus if the hypothesis $H_0$: the CS upper bound is $o(n^{-1/2})$ is rejected by our test, we will have falsified the analyst's justification for the validity of her Wald CIs. In this work, we exhibit a valid assumption-lean falsification test of $H_0$, without relying on complexity-reducing assumptions on $b,p$, or their estimates $hatb,hatp$. Simulation experiments are conducted to demonstrate how the proposed assumption-lean test can be used in practice. An unavoidable limitation of our methodology is that no assumption-lean test of $H_0$, including ours, can be a consistent test. Thus failure of our test to reject is not meaningful evidence in favor of $H_0$.

Cites work

• A class of U-statistics and asymptotic normality of the number of k- clusters
• Adaptive estimation of nonparametric functionals
• Adjusting for Nonignorable Drop-Out Using Semiparametric Nonresponse Models
• Automatic Debiased Machine Learning of Causal and Structural Effects
• Characterization of parameters with a mixed bias property
• Convergence rates and asymptotic normality for series estimators
• Debiased machine learning of global and local parameters using regularized Riesz representers
• Deep neural networks for estimation and inference
• Discussion of ``On nearly assumption-free tests of nominal confidence interval coverage for causal parameters estimated by machine learning
• Double/debiased machine learning for treatment and structural parameters
• Doubly Robust Estimation in Missing Data and Causal Inference Models
• Efficient estimation of average derivatives in NPIV models: simulation comparisons of neural network estimators
• Efficient Estimation of Models with Conditional Moment Restrictions Containing Unknown Functions
• Estimation of possibly misspecified semiparametric conditional moment restriction models with different conditioning variables
• Heteroscedasticity-Robust Inference in Linear Regression Models With Many Covariates
• Higher order inference on a treatment effect under low regularity conditions
• Higher order influence functions and minimax estimation of nonlinear functionals
• Higher order tangent spaces and influence functions
• Identification of Causal Effects Using Instrumental Variables
• Identifying causal effects with proxy variables of an unmeasured confounder
• Inference in linear regression models with many covariates and heteroscedasticity
• Instrumental Variable Estimation of Nonparametric Models
• Kernel-based semiparametric estimators: small bandwidth asymptotics and bootstrap consistency
• Leave‐Out Estimation of Variance Components
• Maximum Likelihood Specification Testing and Conditional Moment Tests
• Measurement Error Models with Auxiliary Data
• Minimax estimation of a functional on a structured high-dimensional model
• Minimax estimation of the integral of a power of a density
• Nonparametric regression using deep neural networks with ReLU activation function
• On nearly assumption-free tests of nominal confidence interval coverage for causal parameters estimated by machine learning
• Optimal sup-norm rates and uniform inference on nonlinear functionals of nonparametric IV regression
• Rejoinder: On nearly assumption-free tests of nominal confidence interval coverage for causal parameters estimated by machine learning
• Robust inference on average treatment effects with possibly more covariates than observations
• Semiparametric efficiency bounds
• Semiparametric efficiency in GMM models with auxiliary data
• Semiparametric minimax rates
• Sieve semiparametric two-step GMM under weak dependence
• Some new asymptotic theory for least squares series: pointwise and uniform results
• The Asymptotic Variance of Semiparametric Estimators
• The Bayesian analysis of complex, high-dimensional models: can it be CODA?
• The Hardness of Conditional Independence Testing and the Generalised Covariance Measure
• The influence function of semiparametric estimators
• The semiparametric efficiency bound for models of sequential moment restrictions containing unknown functions
• Twicing Kernels and a Small Bias Property of Semiparametric Estimators
• Two-step estimation and inference with possibly many included covariates

This page was built for publication: Assumption-lean falsification tests of rate double-robustness of double-machine-learning estimators

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6199660)