Geodesic Causal Inference
read the original abstract
Adjusting for confounding and imbalance when establishing statistical relationships is an increasingly important task, and causal inference methods have emerged as the most popular tool to achieve this. Existing methodology has been developed primarily for outcomes that lie in Euclidean spaces. We introduce here a general framework for causal inference when outcomes reside in general geodesic metric spaces, where we draw on a novel geodesic calculus that facilitates scalar multiplication for geodesics and the quantification of treatment effects through the concept of geodesic average treatment effect. Using ideas from Fr\'echet regression, we obtain a doubly robust estimation of the geodesic average treatment effect and results on consistency and rates of convergence for the proposed estimators. We also develop an intrinsic uncertainty quantification framework for the treatment effect based on Fr\'echet objective functions. The proposed framework is illustrated through simulations and real data applications, including network-valued outcomes from New York City taxi trips to assess the impact of the COVID-19 pandemic, and compositional data on U.S. state-level energy sources to study the effect of coal mining.
This paper has not been read by Pith yet.
Forward citations
Cited by 5 Pith papers
-
Wasserstein Policy Learning for Distributional Outcomes
Establishes finite-sample regret bounds of order sqrt(N-dim(Π)/N) for IPW and DR estimators in Wasserstein policy learning with distributional outcomes, plus a matching minimax lower bound.
-
IV regression with distribution-valued outcomes
IVFR extends global Fréchet regression to endogenous covariates via projection of IV-weighted quantile curves onto valid distributions in 2-Wasserstein space, with weak convergence to a Gaussian process and valid mult...
-
Shrinkage priors for Bayesian Substitute Confounders
Bayesian shrinkage priors on factor models produce sparse substitute confounders that support consistent regression-adjusted causal estimates under latent variable identification assumptions.
-
SphUnc: Hyperspherical Uncertainty Decomposition and Causal Identification via Information Geometry
SphUnc decomposes uncertainty via hyperspherical von Mises-Fisher latents and performs causal identification through structural models on those latents.
-
A Test for Treatment Heterogeneity under a Distributional Difference-in-Difference Framework
Introduces a nonparametric distributional DiD test that transports control-group drift via optimal transport to build a counterfactual and uses an RKHS MMD statistic to test equality.
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.