Randomization inference with general interference and censoring

Abstract

Interference occurs between individuals when the treatment (or exposure) of one individual affects the outcome of another individual. Previous work on causal inference methods in the presence of interference has focused on the setting where it is a priori assumed that there is “partial interference,” in the sense that individuals can be partitioned into groups wherein there is no interference between individuals in different groups. Bowers et al. (2012, Political Anal, 21, 97–124) and Bowers et al. (2016, Political Anal, 24, 395–403) consider randomization‐based inferential methods that allow for more general interference structures in the context of randomized experiments. In this paper, extensions of Bowers et al. that allow for failure time outcomes subject to right censoring are proposed. Permitting right‐censored outcomes is challenging because standard randomization‐based tests of the null hypothesis of no treatment effect assume that whether an individual is censored does not depend on treatment. The proposed extension of Bowers et al. to allow for censoring entails adapting the method of Wang et al. (2010, Biostatistics, 11, 676–692) for two‐sample survival comparisons in the presence of unequal censoring. The methods are examined via simulation studies and utilized to assess the effects of cholera vaccination in an individually randomized trial of 73 000 children and women in Matlab, Bangladesh.

Estimating social influence in a social network using potential outcomes.

“…Social influence effects that originate from “first-order” interactions (with neighbors), and “second-order” interactions (with neighbors of neighbors), may be posited using the causal model: where ui=(∑j,k=1NAijAjk)−1∑j,k=1NAijAjkxk is the proportion of neighbors-of-neighbors of individual i assigned to treatment. We refer readers to Bowers et al (2012); Sussman and Airoldi (2017), and Loh, Hudgens, Clemens, Ali, and Emch (2019) for other causal models.…”

“…where u i ϭ ͑ ͚j,kϭ1 N A ij A jk ͒ Ϫ1 ͚j,kϭ1 N A ij A jk x k is the proportion of neighbors-of-neighbors of individual i assigned to treatment. We refer readers to Bowers et al (2012); Sussman andAiroldi (2017), andLoh, Hudgens, Clemens, Ali, andEmch (2019) for other causal models.…”

Methods for Assessing Spillover in Network-Based Studies of HIV/AIDS Prevention among People Who Use Drugs

“…In this randomized design, one approach to estimate spillover effects is a Horvitz–Thompson type estimator with unequal probability sampling with inverse probability weighted (IPW) estimators, and conservative variance estimators have been proposed [ 64 ]. Although randomization provides protection against confounding, randomized designs can be vulnerable to other issues, including generalizability [ 91 ], measurement error of spillover sets [ 92 , 93 ], non-compliance [ 84 ], and selection bias due to differential loss to follow-up [ 94 , 95 ], that warrant further consideration when evaluating spillover (see Section 3.5.2 ).…”

Causal Effects with Hidden Treatment Diffusion on Observed or Partially Observed Networks

“…The causal inference literature has developed a number of statistical methods to deal with interference (Aronow and Samii, 2017;Forastiere et al, 2020a;Papadogeorgou et al, 2019;Aronow et al, 2019;Miles et al, 2019;Loh et al, 2020;Tortù et al, 2020;Leung, 2020a). In particular, some recent studies have proposed estimators for treatment and spillover effects under the assumption of on partial (or clustered) interference (Sobel, 2006;Hudgens and Halloran, 2008;Tchetgen and VanderWeele, 2012;Liu and Hudgens, 2014;Liu et al, 2016;Kang and Imbens, 2016;Forastiere et al, 2016;Basse and Feller, 2018;Forastiere et al, 2019a), where units are clustered in exogenous groups and spillover mechanisms are assumed to occur only within groups.…”