A Recipe for inferference: Start with Causal Inference. Add Interference. Mix Well with R

Area of Science:

• Causal inference
• Statistical modeling
• Epidemiology
• Political science

Background:

• Interference in causal inference occurs when one subject's treatment affects others, potentially distorting results.
• Inverse probability weighted (IPW) estimators are standard for causal effects without interference.
• Existing IPW methods have been extended to account for interference within groups.

Purpose of the Study:

• To introduce inferference, an R package designed to compute IPW causal effect estimates in the presence of group-level interference.
• To provide a practical tool for researchers dealing with complex causal relationships influenced by spillover effects.

Main Methods:

• Utilizes inverse probability weighted (IPW) methods extended to handle interference within defined groups.
• Implements algorithms for estimating direct and indirect (spillover) causal effects.
• Demonstrates package functionality using real-world examples.

Main Results:

• The inferference package provides a computational solution for estimating causal effects with interference.
• The package facilitates the application of advanced IPW methods to observational data.
• Successful illustration of the package in political science and infectious disease contexts.

Conclusions:

• The inferference R package offers a valuable tool for causal inference research where interference is a concern.
• Accurate estimation of causal effects in the presence of interference is crucial for reliable research conclusions.
• The package supports the analysis of direct and indirect effects in group-based settings.