Simple Yet Efficient Estimators For Network Causal Inference...

on YouTube ↗33m

" summary

Shows that under a heterogeneous linear outcomes model the total treatment effect reduces to a population mean of per-node “influences,” so a plain Bernoulli randomized assignment (marginal probability p) plus an inverse‑propensity/Horvitz–Thompson style estimator is unbiased and computationally trivial even without graph knowledge. Key quantitative claim: estimator variance scales like D^2/(p n) (D = node degree), and extensions cover low‑degree polynomials or the unknown‑baseline case (which then needs more graph info).

AI-written summary · about

" chapters (10)

1. 0:00Intro
2. 1:45Randomized Experiment (A/B Testing)
3. 4:46Challenge of Network Interference
4. 7:07Potential Outcomes Model
5. 8:46Simple first attempt
6. 12:02Brief Literature Review
7. 16:14Preview of Result
8. 18:02Heterogeneous Linear Outcomes Model
9. 21:37Total Treatment Effect with baseline estimates
10. 24:29Reduction to estimating population mean

" related talks

• 30mCausal Estimands And InterventionsVanessa Didelez · Learning from Interventions 2022
• 49mCausal Inference in Complex Systems: Network Interference, Strategic Agents, and BeyondPanos Toulis · Quantifying Uncertainty: Stochastic, Adversarial, and Beyond
• 01h 11mStatistical Limits of Causal InferenceSivaraman Balakrishnan · Modern Paradigms in Generalization Boot Camp 2024
• 33mExplainable causal inference resultsThomas Nägele · PyData Eindhoven 2021
• 57mA New Perspective on High-Dimensional Causal InferencePragya Sur · Deep Learning Theory Workshop and Summer School 2022
• 30mCounterfactual Inference For Sequential Experiment DesignRaaz Dwivedi · Learning from Interventions 2022
• 31mLearning And Testing Causal Models: A Property Testing ViewpointArnab Bhattacharyya · Learning from Interventions 2022
• 24mA Friendly Introduction To CausalityAlex Dimakis · MLconf SF 2016