Rubin's Causal Model

Rubin’s Causal Model is a statistical framework that helps us understand cause and effect between variables based on the framework of potential outcomes

• potential outcomes: 2 possible outcomes that could occur to a unit. We can turn observable outcomes into observed and counterfactual using switching equation
• treatment assignment: the investigator can (at least in principle) assign at random whether a unit receives treatment or not. *Each subject’s response is assumed to depend only on their assignment
• replication: at least one unit is assigned to treatment and at least one unit is assigned to control
• assignment mechanism is ignorable: the treatment is assigned independent of potential outcomes. We can ignore what would happen if the units receive the treatment or control.
• SUTVA:

  The assumption in Rubin’s Causal Model that the response on one unit should be unaffected by the assignment of treatments to the other units, no matter what treatments other units receive. In other words 1) there is no inference across units and 2) treatment means the same thing for every unit. This assumption requires us to be clear about units, treatments, and potential outcomes.

  Link to original

  Preferring RCT or quasi-experiments because everything mentioned above is clear. causation in observational studies is very limited, but as long as we can imagine and calculate counterfactual

Guide

1. What are the units
2. What are the treatments
3. What are the potential outcomes
4. What is the assignment mechanism
5. Is this assignment mechanism useful for causal inference
6. Would it have helped if we isolated a confounder

We can calculate the average treatment effect

Usage

While the Rubin’s Causal Model primarily focuses on estimating the effects of causes (individual treatment effect), it can also be used to some extent to understand what specific causes contributed to observed outcomes. However, the primary strength of the RCM lies in estimating treatment effects and conducting causal inference to understand the impact of interventions or treatments on outcomes.