On causal inference in the presence of interference

Area of Science:

• Statistics
• Causal Inference
• Social Network Analysis

Background:

• Interference arises in studies where individual outcomes are influenced by others' treatments or exposures.
• Social interactions are a common source of interference, complicating standard causal inference methods.
• Existing literature on causal inference with interference is developing but requires further advancement.

Purpose of the Study:

• To summarize existing concepts and results in causal inference with interference.
• To extend the literature by presenting new findings for finite sample inference.
• To introduce novel inverse probability weighting estimators and causal estimands relevant to interference.

Main Methods:

• Review and synthesis of current literature on causal inference with interference.
• Development of new theoretical results for finite sample inference in the presence of interference.
• Formulation of novel inverse probability weighting (IPW) estimators tailored for interference settings.
• Definition and exploration of new causal estimands to capture complex relational effects.

Main Results:

• The study provides a comprehensive overview of interference in causal inference.
• New theoretical results are presented for statistical inference with limited sample sizes under interference.
• Novel IPW estimators are proposed to address challenges posed by interference.
• New causal estimands are introduced to better understand treatment effects in socially connected groups.

Conclusions:

• Causal inference in the presence of interference presents unique challenges that necessitate specialized methods.
• This work contributes new tools and theoretical insights for analyzing data with interference.
• The proposed methods and estimands offer advancements for researchers studying social interactions and their impact on outcomes.