Causal Inference Basics
Estimate causal effects from observational data and state your assumptions out loud
A taste of a lesson
Estimating a loyalty programme's effect on spend. Should I control for 'number of store visits' in my regression?
Probably not, if the programme changes how often people visit. Then visits are a mediator: part of the path from programme to spend. Controlling for them removes the effect that flows through extra visits, so you would underestimate the total effect. If you want visits held fixed, that is a different, direct effect question. What you should adjust for are things that influence both joining and spending before joining, such as past spend. Can you list three pre programme variables that might drive both?
Written by the teacher as an example. In your lesson the tutor answers your own questions, and like any AI it can be wrong.
What you will be able to do
- Define a causal estimand using potential outcomes
- Draw a causal diagram and choose an adjustment set
- Apply matching, difference in differences, regression discontinuity and instruments in concept
- State and probe the key assumption behind each method
- Report a causal estimate with sensitivity analysis and caveats
Lesson plan
- 1 Potential outcomes Frame causal questions as comparisons of what would happen with and without treatment. Start
- 2 Causal diagrams Draw a DAG and decide which variables to adjust for. Start
- 3 Matching and propensity scores Balance treated and control groups on observed covariates. Start
- 4 Difference in differences Estimate effects from before and after comparisons with a control group. Start
- 5 Discontinuities and instruments Use cutoffs and instruments to mimic randomisation. Start
- 6 Sensitivity and reporting Probe robustness and present causal results honestly. Start
Try asking
About this tutor
An advanced tutor for analysts and data scientists who need to answer 'what was the effect?' when a randomised experiment is not possible. You will learn potential outcomes, draw causal diagrams to choose what to adjust for, and work through matching and propensity scores, difference in differences, regression discontinuity and instrumental variables, always with the key assumption of each method in plain words. Lessons end with sensitivity analysis and how to report causal estimates with honest caveats. Light algebra is used; the focus is on design and judgement rather than software.
Reviews
Students can review a tutor after a paid lesson. Nobody has yet.
About the teacher
Experiments, causal questions and responsible models, explained for decision makers
9 tutors 437 lessons taught Sample
I help analysts and product people answer the question behind most data work: did this change cause that result? I teach A/B testing, experiment design and the basics of causal inference, plus the responsible side of modelling: fairness checks and explaining predictions. My background is in product analytics and experimentation, where I learned that a clear sentence to a decision...
See Kojo's profile and tutorsMore like this
Other tutors on the same or nearby topics.