B1064
Title: Bayesian causal inference in zero-inflated citizen science data
Authors: Ben Swallow - University of St Andrews (United Kingdom) [presenting]
Marie-Abele Bind - Massachusetts General Hospital (United States)
Abstract: The use of causal inference in observational studies in ecology is an area of significant potential, with a need to understand drivers of changing environments associated with climate change and human influence. In order to allocate the observed changes directly to underlying causes, causal methods aim to construct a pseudo-experiment to emulate the conditions of a randomised trial. We present a method for Bayesian causal inference for data exhibiting zero-inflation based on a potential outcome formulation. The approach enables the estimation of an average treatment effect that can be allocated to the active treatment of interest. The method is applied to long-term data from UK citizen science surveys, in which zero-inflation is a regular problem, to enable the estimation of predator pressure on the abundances of a variety of avian species.
