Building a Simple Causal DAG
You'll learn how to translate domain knowledge into a visual causal diagram by identifying causes, effects, and their directional relationships.
From Domain Knowledge to Visual Structure
Now that you understand what DAGs are (from lesson 1468), it's time to build one yourself. Creating a causal DAG means drawing arrows from causes to their effects based on what you know about how the world works—not from data patterns, but from substantive knowledge.
The process is straightforward:
- List relevant variables in your research question
- Identify direct causal relationships (does X directly cause Y?)
- Draw arrows from causes to effects
- Check for cycles (remove them—DAGs are acyclic!)
- Document your assumptions (why did you draw each arrow?)
A Simple Example: Coffee and Productivity
Imagine studying whether coffee affects work productivity. Your DAG might include:
- Coffee → Alertness (coffee directly increases alertness)
- Alertness → Productivity (alertness directly boosts productivity)
- Sleep Quality → Alertness (poor sleep reduces alertness)
- Sleep Quality → Coffee (poor sleep makes you drink more coffee)
This creates a simple DAG where Sleep Quality is a confounder (affecting both coffee consumption and productivity through alertness). Notice how each arrow represents a causal belief, not just a correlation you observed in data.
Why This Matters
Your DAG encodes your assumptions visually. It helps you identify confounders (like Sleep Quality), mediators (like Alertness), and which variables you need to control for. If your assumptions are wrong, your causal conclusions will be wrong—but at least the DAG makes those assumptions explicit and debatable.
Key Takeaway: Building a causal DAG means translating domain expertise into a directed graph where arrows represent direct causal effects, making your assumptions transparent and testable.