HarvardX: Causal Diagrams: Draw Your Assumptions Before Your Conclusions course Syllabus

Overview: This course introduces the foundational concepts of causal reasoning using causal diagrams, also known as Directed Acyclic Graphs (DAGs). Designed for beginners in data science and research, it emphasizes critical thinking over technical computation. You'll learn to map assumptions, identify sources of bias, and guide data analysis decisions using visual tools. The course spans approximately 8–11 weeks of part-time study, with 4–6 hours per week. It concludes with a practical final project applying causal diagrams to real-world scenarios.

Module 1: Introduction to Causal Thinking

Estimated time: 6 hours

• Why causal reasoning matters in data analysis
• Distinguishing between association and causation
• Real-world examples of misleading correlations
• Common pitfalls in interpreting observational data

Module 2: Building Causal Diagrams (DAGs)

Estimated time: 8 hours

• Components of DAGs: nodes, arrows, and directionality
• Rules for constructing valid causal diagrams
• Translating assumptions into graphical models
• Practicing DAGs for common research questions

Module 3: Confounding, Mediation, and Bias

Estimated time: 10 hours

• Identifying confounders and their impact on estimates
• Understanding mediation and when not to adjust
• Collider bias and selection effects
• Avoiding common mistakes in variable adjustment

Module 4: Using Causal Diagrams in Analysis

Estimated time: 10 hours

• How DAGs guide variable selection
• Strategies for correct adjustment in regression
• Estimating direct and total causal effects
• Applying DAGs to observational study designs

Module 5: Strengthening Causal Inference

Estimated time: 8 hours

• Evaluating alternative causal structures
• Sensitivity analysis using DAGs
• Integrating domain knowledge with diagrams
• Improving credibility of data-driven conclusions

Module 6: Final Project

Estimated time: 10 hours

• Construct a DAG for a real-world research question
• Identify key sources of bias and confounding
• Justify analytical choices based on the diagram

Prerequisites

• Basic understanding of data analysis concepts
• Familiarity with research or policy contexts
• No coding experience required

What You'll Be Able to Do After

• Explain why correlation does not imply causation
• Construct and interpret causal diagrams (DAGs)
• Identify confounders, mediators, and colliders
• Guide data analysis using causal assumptions
• Improve study design and avoid biased conclusions