Machine Learning and AI Foundations: Causal Inference and Modeling

Learn about the modeling techniques and experimental designs that allow you to establish causal inference, and how to use them.

Introduction

• Thinking about causality
• What you should know

1. Experimental Design and Statistical Controls

• The investigator, the jury, and the judge
• Fisher and experiments
• John Snow and natural experiments
• Double blind studies
• Control variables (ANCOVA)
• Judea Pearl: Problems with control variables
• Moderation, mediation, and lurking variables
• Simpson's paradox
• Challenge: Moderation, mediation, or a third variable
• Solution: Moderation, mediation, or a third variable

2. Conditional Probability and Bayes' Theorem

• Turing, Enigma, and CAPTCHA
• Enigma and uncertainty
• Developing an intuition for Bayes with Wordle
• Wordle and conditional probability
• Wordle, bans, and bits
• Wordle and Bayes' theorem
• Challenge: Conditional probability and Bayes' theorem
• Solution: Conditional probability and Bayes' theorem

3. Prediction and Proof with Bayesian statistics

• Contrasting frequentist statistics and Bayesian statistics
• Bayesian T-Test with JASP
• Google Optimize
• Bayes and rare events
• Challenge: JASP
• Solution: JASP

4. Causal Modeling with Structural Equation Modeling (SEM)

• Sewell Wright
• Introducing path analysis and SEM
• SEM example: Intention
• Myths about SEM
• Latent variables in SEM
• Finding direction of causality with SEM (PSAT)

5. Causal Modeling with Bayesian Networks

• Judea Pearl and the causal revolution
• Downloading BayesiaLab and resources
• Introducing BayesiaLab: Hair and eye color
• Introduction to causal modeling with Bayesian networks
• Bayesian Networks: Black Swan case study

Conclusion

• Taking causality further