JRSSB Nov 17, 2026

Regression discontinuity designs for functional data and random objects in geodesic spaces

Authors
Daisuke Kurisu Yidong Zhou Taisuke Otsu Hans-Georg Müller
Research Topics
Machine Learning
Paper Information
  • Journal:
    Journal of the Royal Statistical Society Series B
  • DOI:
    10.1093/jrsssb/qkag143
  • Published:
    November 17, 2026
  • Added to Tracker:
    Nov 20, 2026
Abstract

Abstract Regression discontinuity designs (RDDs) are widely used for causal inference in observational studies with cutoff-based treatment assignment, primarily for Euclidean outcomes. We propose the geodesic regression discontinuity design (GRDD), which extends RDDs to complex non-Euclidean outcomes, including networks, compositional data, functional data, and other random objects in geodesic metric spaces. Since algebraic operations are unavailable in such spaces, we define the causal effect at the cutoff as the geodesic connecting the local Fréchet means of untreated and treated outcomes, recovering the classical local average treatment effect in the scalar case. Estimation is conducted intrinsically via local Fréchet regression to preserve geometric validity and interpretability. For inference, we adopt an extrinsic approach by embedding the metric space into a Hilbert space, enabling tractable asymptotic analysis. We establish asymptotic normality and develop bootstrap-based procedures for hypothesis testing and confidence intervals for the treatment effect magnitude. We also propose a data-adaptive bandwidth selection method tailored to RDDs in metric spaces and study its empirical performance. Applications include compositional voting outcomes in UK elections and daily CO concentration curves after the Taipei metro introduction, and we extend the framework to fuzzy designs with imperfect compliance.

Author Details
Daisuke Kurisu
Author
Yidong Zhou
Author
Taisuke Otsu
Author
Hans-Georg Müller
Author
Research Topics & Keywords
Machine Learning
Research Area
Citation Information
APA Format
Daisuke Kurisu , Yidong Zhou , Taisuke Otsu & Hans-Georg Müller (2026) . Regression discontinuity designs for functional data and random objects in geodesic spaces. Journal of the Royal Statistical Society Series B , 10.1093/jrsssb/qkag143.
BibTeX Format
@article{paper1930,
  title = { Regression discontinuity designs for functional data and random objects in geodesic spaces },
  author = { Daisuke Kurisu and Yidong Zhou and Taisuke Otsu and Hans-Georg Müller },
  journal = { Journal of the Royal Statistical Society Series B },
  year = { 2026 },
  doi = { 10.1093/jrsssb/qkag143 },
  url = { https://doi.org/10.1093/jrsssb/qkag143 }
}