victorhu95 / Causal-Inference

Causal-Inference

Table of Contents

1. Week 2 Overview
2. Two Main Ways to Study Cause and Effect
3. The Classical Paradigm
   • Bradford Hill Criteria
   • Applying the Classical Approach
   • Example: Air Pollution and Health
4. Potential Outcomes Paradigm
   • Rubin Causal Model
5. Traditional Statistics vs. Causal Inference
6. Potential Outcomes vs. Classical Paradigms
7. Foundational Concepts
   • Units
   • Treatment
   • Causes
   • Potential Outcomes

Week 2 Overview

Two Main Ways to Study Cause and Effect

1. The Classical Paradigm
2. Potential Outcomes Paradigm

The Classical Paradigm

Bradford Hill Criteria

1. Strength
2. Consistency
3. Specificity
4. Timing
5. Gradient
6. Plausibility
7. Coherence
8. Analogy
9. Experimental Evidence

Applying the Classical Approach

1. Observation
2. Gather More Information
3. Apply Bradford Hill Criteria
4. Make a Judgment

Example: Air Pollution and Health

1. Gathering Evidence
2. Review by Authorities
3. Judging the Evidence
4. Final Statement

Potential Outcomes Paradigm

Rubin Causal Model

1. Think of Possible Outcomes
2. How Are Treatments Given
3. Model for The Science

Traditional Statistics vs. Causal Inference

• What Statistics Usually Does: Associational Inference
• What Causal Inference Usually Does

Potential Outcomes vs. Classical Paradigms

1. What is the Cause?
2. What is the Effect?

Foundational Concepts

Units

• Objects of Study
• Statistical Population
• Key Variables

Treatment

• Definition
• Thinking Experimentally
• Key Feature
• Contrast with Attributes
• Role of Time

Causes

1. Experimental Treatments as Causes
2. Relative Effects
3. Not Limited to Randomized Studies

Potential Outcomes

• (Y^t_i), (Y^c_i), (Y^Z_i)
• Innate Characteristics Before Treatment
• Assignment to (Z) Determines Observed Outcome
• Relationship Between Potential Outcomes and Observed Data