A0558
Title: Statistical inference in high-dimensional regression with hidden confounders by double debiased LASSO estimator
Authors: Guoyou Qin - Fudan University (China) [presenting]
Abstract: The statistical inference in the high-dimensional linear regression is considered to have hidden confounders. A double debiased LASSO estimator is proposed based on the spectral transformation and the approximately inverse empirical covariance matrix of the transformed design matrix. The proposed estimator corrects the bias from the estimation of the high-dimensional coefficients and the hidden confounders without the sparse assumption on the precision matrix of the component of covariates unaffected by confounders. The asymptotic properties of the estimator for the individual component and finite-dimensional subset of the coefficient vector are presented. The performance of the estimator is investigated through simulation experiments and a real dataset.
