Causal Inference Working Group: Hyemin Yeon, Causal Inference with Functional Treatments
Speaker: Hyemin Yeon, Graduate Student, Department of Statistics, Iowa State University
Title: Causal inference with functional treatments
Abstract: In this talk, I review two recent papers about causal inference with functional treatments: Zhang, Xue, and Wang (2021) and Tan, Huang, Zhang, and Yin (2022) coded as [ZXW21] and [THZY22] respectively (the only papers about functional treatments to the best of my knowledge). The main issue with functional treatments is to define propensity scores or stablized weights because there is no density for functional data (). Upon defining propensity scores or stablized weights appropriately, the two papers next propose the estimation methods of the stablized weights; [ZXW21] follow while [THZY22] consider a similar approach to the one in . Finally, under linear structure, the two papers estimate the functional causal effect parameter; [ZXW21] apply the naive basis expansion approach while [THZY22] use the regression estimator based on the functional principal component analysis. The methods in the two papers are introduced for each issue and the focus is on comparing how differently the two papers connect causal inference and functional data analysis.