Assumptions in causal inference
Causal inference is fundamentally built upon a foundation of carefully constructed assumptions. Assumptions represent the underlying beliefs about the origins of our data. Often, these assumptions are not directly verifiable by the data itself, which necessitates the need to pre-suppose their existence. Identifying these assumptions is a critical challenge, and this section aims to provide clear guidance on how to do so.
It’s not a surprise anymore that central to causal inference is the task of identifying causal effects. This is distinctly different from the challenges of estimation found in traditional statistics and machine learning. Identification involves determining whether it’s possible to learn a causal effect from the data, based on the underlying assumptions. Once these effects are identified, estimation—common to both causal inference and traditional statistics—aims to quantify the size or nature of these effects...
