Article·causal-learn.readthedocs.io
causal-inferencecausal-discoveryDAGstructural-causal-modelsPC-algorithmGESFCM
Causal Discovery
Learn how to automatically discover causal relationships from observational data using various algorithms, enabling data-driven hypothesis generation.
intermediate2-3 hours5 steps
The play
- Introduction to Causal DiscoveryUnderstand the fundamental concepts of causal discovery, including causal relationships, directed acyclic graphs (DAGs), and the difference between correlation and causation.
- Constraint-Based Algorithms (PC Algorithm)Explore constraint-based algorithms like the PC algorithm. Learn how it uses conditional independence tests to identify potential causal relationships and construct a DAG.
- Score-Based Algorithms (GES)Learn about score-based algorithms like GES (Greedy Equivalence Search). Understand how they use a scoring function (e.g., BIC, MDL) to evaluate different DAG structures and search for the one that best fits the data.
- Functional Causal Models (FCMs)Explore Functional Causal Models (FCMs) and their application in causal discovery. Understand how they model causal relationships as functions and can handle non-linear relationships.
- Evaluating and Interpreting ResultsLearn how to evaluate the results of causal discovery algorithms, including assessing the plausibility of the discovered causal relationships and considering potential biases and limitations.
Starter code
Start by understanding the difference between correlation and causation and the basics of DAGs.
Source