Data Science
Causal Inference
This guide helps you get your bearings in Causal Inference before you start exploring the interactive timeline, framework graph, and concept maps.
Before You Dive In
- Causal Inference asks how to extract reliable insight and decision value from messy, high-dimensional data.
- Rough timeline: classical statistics and sampling -> large-scale data mining -> ML production systems -> causal and responsible data science.
- Start with the distinction between prediction, explanation, and intervention; each needs different methods and assumptions.
- In Noosaga, compare frameworks by objective: forecasting accuracy, causal validity, operational robustness, or interpretability.
Key Terms to Know
GeneralizationModel performance on unseen data drawn from relevant distributions.
Bias-variance tradeoffError decomposition balancing underfitting and overfitting tendencies.
Feature engineeringConstructing informative model inputs from raw data.
Causal inferenceMethods for estimating intervention effects beyond observational correlation.
Data pipelineAutomated flow for ingesting, transforming, validating, and serving data.
Common Confusions
Treating correlation as evidence of actionable causation.
Assuming offline benchmark performance predicts production impact.
Confusing model interpretability with true causal explanation.
Recommended Reading
An Introduction to Statistical Learning— Gareth James et al.
2021The Elements of Statistical Learning— Trevor Hastie, Robert Tibshirani & Jerome Friedman
2009Causal Inference: The Mixtape— Scott Cunningham
2021How to Use the Interactive View
1
Explore the timeline
Open the interactive view and scan the framework timeline. Which frameworks came first? Which ones overlap? Where are the big transitions?
2
Read the articles
Click into individual frameworks to read what each one claims, where it came from, and how it relates to its neighbors.
3
Check the concept map
See how the key ideas within a framework connect. This is useful for figuring out what to learn first and what depends on what.
4
Test yourself
Take the quiz for any framework you've read about. It's a quick way to find out whether you actually understood the core ideas or just skimmed them.