Introduction to Experimental Design

on YouTube ↗01h 01m

" summary

Overview of concrete results and assumptions: experimental design targets causal-structure learning under DAG and no-latents assumptions, using DO/knockout, perfect and soft interventions modeled by an interventional DAG / I-Markov equivalence formalism. Key worst-case counts: if only single-node interventions are allowed it needs P−1 single-node interventions (P≥3, no observational data); allowing arbitrary simultaneous targets cuts this to about floor(log2 P)+1 interventions via a binary-index scheme. All claims assume noiseless identifiability and infinite-sample conditional-invariance tests.

AI-written summary · about

" chapters (8)

1. 0:00Introduction
2. 3:30Problem Setting
3. 12:10Noiseless
4. 14:58knockdown interventions
5. 19:44markov equivalents
6. 22:51graphical characterization
7. 27:23Edges
8. 31:59I essential graphs

" related talks

• 24mA Friendly Introduction To CausalityAlex Dimakis · MLconf SF 2016
• 41mOn the Path to Causal InferenceMark Farragher · PyData London 2019
• 56mWhat can be passively learned about causality?Andrew Lampinen · Understanding Higher-Level Intelligence from AI, Psychology, and Neuroscience Perspectives
• 30mCausal Estimands And InterventionsVanessa Didelez · Learning from Interventions 2022
• 33mLearning Abstract Causal Structure From Causal Interventions In Children And Artificial AgentsAlison Gopnik · Learning from Interventions 2022
• 44mCausal ReductionsJoris Mooij · Quantum Physics and Statistical Causal Models (Simons Institute workshop, 2022)
• 31mLearning And Testing Causal Models: A Property Testing ViewpointArnab Bhattacharyya · Learning from Interventions 2022
• 49mCausal Inference in Complex Systems: Network Interference, Strategic Agents, and BeyondPanos Toulis · Quantifying Uncertainty: Stochastic, Adversarial, and Beyond