Written By: Qingyang Xu (website)

Date Created: July 6, 2022

Last Modified: October 1, 2023

• Applied Causal Inference Powered by ML and AI (Chernozhukov et al., 2022)
• Getting Started with Causal Inference: https://causalinference.gitlab.io/
• Causal Inference with Applications (Imai, Harvard 2023) https://imai.fas.harvard.edu/teaching/index.html

Causal Inference with Applications

• Summary of the Harvard course STATS 286 taught by Professor Imai

1. Potential Outcomes

Setup

• Treatment $T_i$ (e.g., whether voter $i$ is canvassed)
• Outcome $Y_i$
• Pre-treatment covariates $X_i$
• Causal effect for voter i: $\tau_i=Y_i(1)-Y_i(0)$

Key Assumptions

• Causal ordering: $T_i \rightarrow Y_i$
• Consistency: $Y_i = Y_i (t)$ whenever $T_i = t$
  • No hidden multiple versions of treatment
  • No hidden different administration of treatment
• No interference between units: $Y_i (T_1, \cdots , T_n) = Y_i(T_i)$