README.md

CATE Example: Conditional Average Treatment Effect

This example demonstrates how to calculate the Conditional Average Treatment Effect (CATE) using the DeepCausality library.

Specifically, it models the effect of a medication on blood pressure for a specific subgroup: patients over 65 years old.

How to Run

From the root of the deep_causality project, run:

cargo run -p classical_causality_examples --example cate_example

---

Key Concepts

What is CATE?

The Conditional Average Treatment Effect is the average causal effect of a treatment for a specific subgroup of the population. Unlike the Average Treatment Effect (ATE), which considers the entire population, CATE focuses on individuals who share certain characteristics.

How It Works

1. Patient Population: A population of patients is created, each with attributes like age and initial blood pressure stored in a BaseContext.

2. Subgroup Selection: Patients over 65 are filtered to form the target subgroup.

3. Counterfactual Contexts: For each patient in the subgroup, two alternate realities are created:

   • Treatment Context: Drug is administered (drug_administered = 1.0)
   • Control Context: No drug (drug_administered = 0.0)

4. Causaloid Evaluation: The same causal logic (drug_effect_logic) is evaluated against both contexts to compute:

   • Y(1): Potential outcome with treatment
   • Y(0): Potential outcome without treatment

5. Individual Treatment Effect (ITE): For each patient: ITE = Y(1) - Y(0)

6. CATE Calculation: The average of all ITEs in the subgroup gives the CATE.

EPP Principle

This example demonstrates the EPP's power of Contextual Alternation - the same causal model can be evaluated against different contexts to simulate potential outcomes, enabling counterfactual reasoning without modifying the underlying causal laws.

Reference

For more information on the EPP, please see chapter 5 in the EPP document: https://github.com/deepcausality-rs/papers/blob/main/effect_propagation_process/epp.pdf