Confounding Equivalence in Causal Inference

TL;DR

This paper introduces a straightforward test to determine if two variable sets in a causal diagram have equivalent bias-reducing capabilities, based on admissibility or Markov boundary conditions, aiding covariate selection.

Contribution

It presents a novel, simple criterion for equivalence of bias reduction potential in causal diagrams, enhancing covariate selection and model testing methods.

Findings

The test is applicable when both sets are admissible or have identical Markov boundaries.

It simplifies the process of covariate selection in causal inference.

The approach aids in model testing for causal diagrams.

Abstract

The paper provides a simple test for deciding, from a given causal diagram, whether two sets of variables have the same bias-reducing potential under adjustment. The test requires that one of the following two conditions holds: either (1) both sets are admissible (i.e., satisfy the back-door criterion) or (2) the Markov boundaries surrounding the manipulated variable(s) are identical in both sets. Applications to covariate selection and model testing are discussed.