Causal Inference Assumptions: What to Check Before You Run the Model
Nasereliver · x · 2026-10-03
A systematic tutorial on the assumptions required for causal inference with observational data. The author's core point: data itself lacks intelligence — you must think before modelling, since identification is what makes observational data mimic an RCT.
- Covers the fundamental problem of causal inference and both Rubin and Pearl frameworks
- Walks through core assumptions: unconfoundedness, the backdoor criterion, good vs. bad controls, positivity, and the modularity assumption
- Explains SUTVA and its violations (spillover), plus consistency
- For panel/time-series settings: parallel trends and no-anticipation assumptions
- Positioned as a practical checklist for cross-sectional, panel, and time-varying confounding scenarios before estimating treatment effects
More from Research
- Schmidhuber cites three decades of papers on formal theories of creativity and curiosity — SchmidhuberAI · 2026-10-03
- Projection sampling: transforming expert data so SFT learns new skills without forgetting — burkov · 2026-10-03
- Grok 4.7 in Cursor yields numerical candidate for open sphere-inspection problem — xiaosun86 · 2026-10-03
- Pairscore: scoring multiple items at once beats independent ranking for LLM uncertainty — yeewhye · 2026-10-03
- willcb: neurosymbolic world models beat single-rollout RL with value models — willcb · 2026-10-03
- 38 AI-for-Science Papers, Six Lessons: From mRNA Stability to Label-Efficient Neutrino Models — bravo_abad · 2026-10-03