Primary Submission Category: Causal Inference and SUTVA/Consistencies Violations
Nonparametric Causal Survival Analysis under Clustered Interference
Authors: Chanhwa Lee, Michael Hudgens,
Presenting Author: Chanhwa Lee*
Interference arises when a unit’s treatment affects the outcome of other units. Sometimes, units are grouped into clusters, where it is reasonable to assume interference only occurs within cluster, i.e., clustered interference. Several methods exist for estimating various causal estimands under clustered interference from observational data, but either (i) the estimands lack real-world relevance, (ii) the estimators rely on parametric models, and/or (iii) the methods do not accommodate right-censored outcomes. To address these issues, we introduce a general framework for estimating treatment effects in the presence of clustered interference and right censoring. Our method is applicable
to any stochastic policy which modifies the propensity score distribution and thus relevant across diverse settings. Nonparametric sample splitting estimators are constructed, allowing for flexible data-adaptive estimation of nuisance functions, and are consistent and asymptotically normal, converging at the usual parametric rate. Simulation studies demonstrate the finite sample performance of the proposed estimators, and the method is applied to a cholera vaccine study in Bangladesh.