Causally Sound Priors for Binary Experiments

TL;DR

This paper introduces the BREASE Bayesian framework for analyzing binary randomized controlled trials, providing interpretable priors, analytical formulas, and efficient sampling methods to improve causal inference and sensitivity analysis.

Contribution

It proposes a novel prior distribution for binary treatment and outcome analysis, enabling better prior elicitation, dependence modeling, and computational efficiency in causal inference.

Findings

Provides analytical formulas for marginal likelihood and Bayes factors.

Offers an exact posterior sampling algorithm and a fast Gibbs sampler.

Demonstrates improved estimation and hypothesis testing in empirical examples.

Abstract

We introduce the BREASE framework for the Bayesian analysis of randomized controlled trials with a binary treatment and a binary outcome. Approaching the problem from a causal inference perspective, we propose parameterizing the likelihood in terms of the baselinerisk, efficacy, and adverse side effects of the treatment, along with a flexible, yet intuitive and tractable jointly independent beta prior distribution on these parameters, which we show to be a generalization of the Dirichlet prior for the joint distribution of potential outcomes. Our approach has a number of desirable characteristics when compared to current mainstream alternatives: (i) it naturally induces prior dependence between expected outcomes in the treatment and control groups; (ii) as the baseline risk, efficacy and risk of adverse side effects are quantities commonly present in the clinicians' vocabulary, the hyperparameters of the prior are directly interpretable, thus facilitating the elicitation of prior knowledge and sensitivity analysis; and (iii) we provide analytical formulae for the marginal likelihood, Bayes factor, and other posterior quantities, as well as an exact posterior sampling algorithm and an accurate and fast data-augmented Gibbs sampler in cases where traditional MCMC fails. Empirical examples demonstrate the utility of our methods for estimation, hypothesis testing, and sensitivity analysis of treatment effects.