Auto-G-Computation of Causal Effects on a Network

Abstract

Methods for inferring average causal effects have traditionally relied on two key assumptions:(i) the intervention received by one unit cannot causally influence the outcome of another; and (ii) units can be organized into non-overlapping groups such that outcomes of units in separate groups are independent. In this paper, we develop new statistical methods for causal inference based on a single realization of a network of connected units for which neither assumption (i) nor (ii) holds. The proposed approach allows both for arbitrary forms of interference, whereby the outcome of a unit may depend on interventions received by other units with whom a network path through connected units exists; and long range dependence, whereby outcomes for any two units likewise connected by a path in the network may be dependent. Under network versions of consistency and no unobserved confounding, inference is made tractable by an assumption that the networks outcome, treatment and covariate vectors are a single realization of a certain chain graph model. This assumption allows inferences about various network causal effects via the auto-g-computation algorithm, a network generalization of Robins' well-known g-computation algorithm previously described for causal inference under assumptions (i) and (ii).

Causal Inference, Social Networks and Chain Graphs

“…Although our Supreme Court example afforded multiple, ostensibly independent and identically distributed observations from the same chain graph, statistical inference for a single realization of a large chain graph was developed in Tchetgen Tchetgen et al . (). The chain graph model can represent only the true data‐generating distribution if the outcomes are in very specific kinds of equilibria, which may be plausible if the outcomes represent collective beliefs or decisions (as in the Supreme Court example) but are often implausible.…”

“…Tchetgen Tchetgen et al . (), who assumed a conventional causal model for data drawn from units in a network, used chain graph Markov assumptions as a model on the observed data to make inference possible in the full interference setting, where the effective sample size is 1. Sherman and Shpitser () gave a complete identification theory for causal effects in the presence of both network dependence and unobserved confounding in latent variable chain graph models, whereas Bhattacharya et al .…”

Does deforestation drive visceral leishmaniasis transmission? A causal analysis

“…Causal mechanisms have been studied within the framework of causal inference in the presence of interference, allowing for partial interference between units [28][29][30]. More recently, methods have been developed to allow for full interference by assuming that the data can be described as a network [26,[31][32][33], and representing the data as chain graphs [47,48] that allow inferences about the parameters of the joint distribution of the observed data. A chain graph is a mixed graph that enables a network to be constructed from both the directed relationship of the variables (as in figure 2), and the undirected relationship of the study units (i.e.…”

“…infection), the so-called dependent happenings. Only fairly recently methods to allow for such dependence in causal inference contexts became available [26][27][28][29][30][31][32][33].…”

Spillover Effects in the Presence of Unobserved Networks

“…Our paper builds on a growing literature on spillover effects (e.g., Hong and Raudenbush 2006; Sobel 2006; Rosenbaum 2007; Hudgens and Halloran 2008; Tchetgen Tchetgen and VanderWeele 2010), especially spillover effects in networks (e.g., Aronow 2012; Bowers et al 2013; Toulis and Kao 2013; Liu and Hudgens 2014; Ogburn and VanderWeele 2014; Athey et al 2018; Forastiere et al 2016; Aronow and Samii 2017; Eckles et al 2017; Tchetgen Tchetgen et al 2017; Bowers et al 2018). See Halloran and Hudgens (2016) for a recent review about spillover effects in general.…”