How to Create a DAG: An 8-Step Guide to Causal Diagrams
How do you construct a directed acyclic graph (DAG) when all you have is a dataset and a list of variables? In this video, I walk through an eight-step strategy for building a causal DAG from the ground up. We begin by identifying the treatment and outcome, then gradually add the variables that may create confounding, mediation, collider bias, and selection bias. You’ll learn how to: • Identify the treatment and outcome variables • Start with a simple “baby DAG” • Find variables that may cause treatment exposure • Distinguish confounders from variables that only predict treatment • Identify mediators and decide whether controlling for them answers your research question • Recognize colliders and selection variables • Evaluate the assumptions represented by missing arrows • Keep your first DAG simple, defensible, and open to revision Along the way, I use examples involving education and income, exercise and depression, muscle growth, mindfulness, social support, health, and study participation. The central idea is that a DAG is not an objective picture discovered inside your dataset. It is a transparent representation of your causal assumptions. Every arrow makes a theoretical claim—and every missing arrow makes one too. This video is part of my series on DAGs and causal inference. See the links below for the previous video, the complete DAG course, and my other statistics courses. Note that the DAG class is still in progress and I anticipate several more videos/quizzes will be made available in the coming weeks. DAG class: https://simplistics.net/simplistics-c... For the self-guided simplistics course: https://simplistics.net/course/univar... For the self-guided Mixed Models course: https://simplistics.net/course/mixed/ For the self-guided visualization course: https://simplistics.net/course/random... For the self-guided R course: https://simplistics.net/course/introd... For other classes, see: https://simplistics.net/all-courses/ For consulting, see: https://simplistics.net/product/stati...

Causal Inference with DAGs

The fragile foundation of inferential statistics

Unfortunately, You Need to Know What the Jevons Paradox is

The Scariest Chart in Electrical Engineering

Generalized Linear Mixed Models for Everything

Can p-values read your mind? How about Bayesian Statistics?

Academia versus Industry for a Statistics Career

Germans DO Have a Sense of Humour - Jimmy Carr Meets a Frankfurt Audience
![The CEO Who Got Rid of the Managers | Bill Anderson [ARC 2026]](https://i.ytimg.com/vi/O_al-CtL72s/hqdefault.jpg?sqp=-oaymwEjCNACELwBSFryq4qpAxUIARUAAAAAGAElAADIQj0AgKJDeAE=&rs=AOn4CLB7cyfIYdDvUQIdMP-n6cD3CWKxag)
The CEO Who Got Rid of the Managers | Bill Anderson [ARC 2026]

ARIMA Models in R

Science has a problem with causality

General relativity from first principles – Adam Brown

The Psychological Reasons Statistics Training Is So Bad

What economists get wrong about government debt | Prof. Steve Keen

The Strangest Things that Correlate with IQ

Don't Hang Up On AI Scammers. Do THIS Instead.
![Your Phone Is Destroying Your Sense of Meaning | Arthur Brooks [ARC 2026]](https://i.ytimg.com/vi/PfTcgYwW14E/hqdefault.jpg?sqp=-oaymwEjCNACELwBSFryq4qpAxUIARUAAAAAGAElAADIQj0AgKJDeAE=&rs=AOn4CLBzBqlOzXKaS_MJrH8pIKehA8ZKPQ)
Your Phone Is Destroying Your Sense of Meaning | Arthur Brooks [ARC 2026]

Last Lecture Series: “How to Win Without Crushing Your Soul” - Graham Weaver

What Happens to Your Debt if the Dollar Collapses: Top Economist

