Positivity-free Policy Learning with Observational Data - Archive ouverte HAL
Conference Papers Year : 2024

Positivity-free Policy Learning with Observational Data

Abstract

Policy learning utilizing observational data is pivotal across various domains, with the objective of learning the optimal treatment assignment policy while adhering to specific constraints such as fairness, budget, and simplicity. This study introduces a novel positivity-free (stochastic) policy learning framework designed to address the challenges posed by the impracticality of the positivity assumption in real-world scenarios. This framework leverages incremental propensity score policies to adjust propensity score values instead of assigning fixed values to treatments. We characterize these incremental propensity score policies and establish identification conditions, employing semiparametric efficiency theory to propose efficient estimators capable of achieving rapid convergence rates, even when integrated with advanced machine learning algorithms. This paper provides a thorough exploration of the theoretical guarantees associated with policy learning and validates the proposed framework's finite-sample performance through comprehensive numerical experiments, ensuring the identification of causal effects from observational data is both robust and reliable.
Fichier principal
Vignette du fichier
positivity_free_policy_learning.pdf (379.72 Ko) Télécharger le fichier
Origin Files produced by the author(s)

Dates and versions

hal-04236430 , version 1 (10-10-2023)

Licence

Identifiers

  • HAL Id : hal-04236430 , version 1

Cite

Pan Zhao, Antoine Chambaz, Julie Josse, Shu Yang. Positivity-free Policy Learning with Observational Data. AISTATS 2024 - The 27th International Conference on Artificial Intelligence and Statistics, May 2024, Valence, Spain. pp.1918-1926. ⟨hal-04236430⟩
155 View
109 Download

Share

More