Proximal Causal
Proximal causal learning addresses causal inference challenges arising from unobserved confounding variables by leveraging observable "proxy" variables that are related to the unobserved confounders. Current research focuses on developing robust estimation methods, particularly doubly robust estimators and kernel-based approaches, for various treatment types (binary and continuous) and data modalities (including text data). This framework is proving valuable in diverse applications, such as recommendation systems and domain adaptation, by enabling more accurate causal effect estimation even when complete confounder information is unavailable.
Papers
March 12, 2024
January 12, 2024
September 22, 2023
February 10, 2023
January 26, 2023
October 29, 2022