Estimating treatment effects from observational spatial data using a difference-in-differences approach


For most problems involving some kind of spatial exposure, e.g. to a neighborhood environment, it is either unethical, impractical, or impossible to randomize people to different neighborhoods. So, what happens when high- or low-risk individuals are overrepresented in one group vs. another?

Mar 12, 2021 12:00 AM
Jon Zelner
Jon Zelner
Assistant Professor

My research interests include distributed robotics, mobile computing and programmable matter.