Model-free causal inference of binary experimental data

TL;DR

This paper develops model-free, randomization-based methods for causal inference in binary experiments, enabling likelihood and Bayesian inference without modeling assumptions, and introduces sensitivity analysis for potential outcome associations.

Contribution

It introduces novel model-free inferential procedures and sensitivity analysis techniques for binary experimental data, avoiding reliance on super population models.

Findings

Likelihood and Bayesian estimators based on physical randomization

Estimators that only provide estimates within feasible support

Sensitivity analysis characterizing impact of potential outcome association

Abstract

For binary experimental data, we discuss randomization-based inferential procedures that do not need to invoke any modeling assumptions. We also introduce methods for likelihood and Bayesian inference based solely on the physical randomization without any hypothetical super population assumptions about the potential outcomes. These estimators have some properties superior to moment-based ones such as only giving estimates in regions of feasible support. Due to the lack of identification of the causal model, we also propose a sensitivity analysis approach which allows for the characterization of the impact of the association between the potential outcomes on statistical inference.