CausalImages: An R Package for Causal Inference with Earth Observation, Bio-medical, and Social Science Images

Satellite image data represents an emerging resource for research in global development and earth observation, yet no R package currently exists to handle images for causal inference up to now. By causal inference, we refer to the rich literature in statistics (Imbens and Rubin 2016), computer science (Pearl 2009), and beyond (Hernan and Robins 2020). Satellites generate temporally-rich worldwide coverage, capturing the entire Earth's surface at regular intervals, except when obscured by clouds (Burke et al. 2021). Historical archives date back to the 1970s. Unlike snapshots of political, economic, or educational systems at a single time point, satellites revisit each location every 2 weeks or more, providing approximately 26 temporal observations annually. This time-series information has proven valuable for studying phenomena like transportation network growth (Nagne and Gawali 2013), urbanization (Schneider et al. 2009), health and living conditions (Daoud et al. 2023; Chi et al. 2022), living standards (Yeh et al. 2020; Pettersson et al. 2023), and neighborhood characteristics