Randomization inference with general interference and censoring.

Area of Science:

• Causal inference
• Biostatistics
• Epidemiology

Background:

• Interference in experiments occurs when one individual's treatment affects another's outcome.
• Prior methods often assumed limited 'partial interference' between groups.
• General interference structures require advanced statistical approaches.

Purpose of the Study:

• To extend existing causal inference methods for general interference structures.
• To accommodate failure time outcomes subject to right censoring.
• To address challenges in randomization-based tests with treatment-dependent censoring.

Main Methods:

• Extending the Bowers et al. framework for general interference.
• Adapting the Wang et al. method for censored survival data.
• Utilizing simulation studies to validate the proposed methods.

Main Results:

• The proposed methods effectively handle interference and right-censored outcomes in randomized trials.
• Simulation studies demonstrate the robustness and accuracy of the new techniques.
• The methods were applied to a large cholera vaccination trial.

Conclusions:

• The developed methods provide a powerful tool for causal inference in complex experimental settings with censored data.
• Accurate analysis of interference is crucial for understanding treatment effects in real-world scenarios.
• This work advances the statistical methodology for analyzing public health interventions.