A1003
Title: Deep orthogonal learner for conditional quantile treatment effect estimation
Authors: Qixian Zhong - Xiamen University (China) [presenting]
Abstract: In recent years, orthogonal statistical learning has been widely recognized for its ability to reduce sensitivity with respect to nuisance parameters to estimate target parameters, making it an important tool in causal inference, particularly in the estimation of the conditional average treatment effect (CATE). However, its application on conditional quantile treatment effect (CQTE), which offers a more expansive view of the treatment effect than CATE, has not yet been explored comprehensively. A novel method is proposed for learning CQTE. This method shares Neyman orthogonal property, which produces CQTE estimators that are insensitive to small perturbations of nuisance functions. The CQTE nonparametrically is first modeled, and deep learning is used to approach it. The convergence rate of the neural network estimator is established, demonstrating that it achieves the minimax optimal rate of convergence (up to a polylogarithmic factor). This highlights deep learning's ability to identify low-dimensional structures in high-dimensional data. Additionally, CQTE linearly is then modeled to facilitate interpretation and statistical inference. It is proven that the corresponding coefficient and CQTE estimators achieve root-n consistency and asymptotic normality, even if the estimators of the nuisance parameters converge at a slower rate. Through empirical evaluation for numerical studies, the superiority of the method is demonstrated compared to competing methods.
