Addressing Spatial Interference in Causal Analysis

• Activity Number: 14
• Type: Topic Contributed
• Date/Time: Sunday, July 31, 2016 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Keith William Zirkle* and David C. Wheeler and Saba W. Masho
• Companies: Virginia Commonwealth University and Virginia Commonwealth University and Virginia Commonwealth University
• Keywords: interference ; stable unit treatment value assumption ; causal inference ; spatial data ; spatial autocorrelation ; causal analysis

Abstract

A popular model for causal inference is based on potential outcomes if study units receive each of the interventions in the study. A fundamental assumption under this framework is no interference; that is, the potential outcomes of one unit are not affected by the intervention on other units. This assumption does not hold in the presence of spatial autocorrelation, where we may expect spillover or diffusion effects based on units' proximity to other units. In this talk, we will review how the literature currently deals with interference and propose a spatiotemporal model to estimate intervention effects based on spatial neighborhood structure. We will then extend the model to a causal inference framework with appropriate modifications. We will present results of applying the model to the Richmond City Youth Violence Surveillance Study, where we estimate intervention effects in census block groups concerning violent incidents by youth offenders (10- to 24-year-olds) after community-based interventions were implemented in three middle schools.