Description
Causal inference is the attempt to go beyond correlation, and draw conclusions that something is being caused by something else. Questions of robust causal inference are unavoidable in health or social science settings, where available data is observational, meaning we only observe one outcome per individual. One common approach to causal inference is to match each treatment unit to an identical control unit with all the same characteristics. Exact matching is not possible for high dimensional datasets. So, for the best possible alternative, we introduce the dame-flame matching library for discrete datasets, and show the matches are high quality and interpretable.