Bayesian estimation of causal effects from observational categorical data

Vera Kvisgaard, Johan Pensar

Keywords: Bayesian networks, causal inference, causal discovery

TL;DR: A bayesian approach to estimate causal effects between categorical variables from non-experimental data, assuming the true underlying causal model is not known.

Abstract:

We develop a scaleable Bayesian method for estimation of all pairwise causal effects in a system from observational data, under the assumption that the underlying causal model is an unknown discrete Bayesian network and that there are no latent confounders.
Specifically, we build upon the the Bayesian IDA (BIDA) and extend this method to the categorical setting. The key-idea is to combine Bayesian estimation of intervention distributions through the so-called backdoor formula with Bayesian model averaging.
The main motivation of the method is to inform future experiments about plausible strong relationships, and we demonstrate by numerical experiments that our Bayesian modeling averaging approach can be highly relevant for this task.

Submission Number: 22