核心概念
Estimating Conditional Average Treatment Effects (CATEs) using Triple/Debiased Lasso methodology.
摘要
This study explores the estimation and statistical inference of Conditional Average Treatment Effects (CATEs) using high-dimensional linear regression models with sparsity. The methodology involves estimating nuisance parameters, regressing covariates on the difference in outcomes with Lasso regularization, and debiasing the bias introduced by the regularization. The proposed Triple/Debiased Lasso (TDL) estimator demonstrates consistency and asymptotic normality.
Directory:
- Introduction
- Investigates CATE estimation in various disciplines.
- Problem Setting
- Introduces potential outcomes and observations.
- TDL for CATE Estimation
- Introduces high-dimensional linear regression models with sparsity.
- Weighted TDL for CATE Estimation with Lasso
- Introduces the proposed Weighted TDL (WTDL) estimator.
- Approximation of the Difference
- Discusses the estimation of the difference using Cross-Fitting.
- Weighted Least Squares with Lasso Regularization
- Introduces the weighted least squares method.
- Debiased Lasso for the WTDL Estimator
- Discusses debiasing the bias introduced by Lasso.
- Summary
- Summarizes the methodology and theoretical properties.
統計資料
"λ ≍ p log(p)/n."
"maxj,d bΣ2j ≤ M 2."
"λ ≍ K0 p log(p)/n."
引述
"We refer to the debiased estimator as the triple/debiased Lasso (TDL), applying both DML and debiased Lasso techniques."