Bounds and Sensitivity Analysis of the Causal Effect Under Outcome-Independent MNAR Confounding

TL;DR

This paper derives assumption-free bounds for causal effects under outcome-independent MNAR confounding and introduces a sensitivity analysis method to evaluate the robustness of these bounds.

Contribution

It provides the first bounds for causal effects with outcome-independent MNAR confounding and offers a sensitivity analysis approach to assess their reliability.

Findings

Derived assumption-free bounds for causal contrasts under MNAR confounding.

Introduced a sensitivity analysis method to evaluate bounds.

Applicable to scenarios with outcome-independent missingness.

Abstract

We report assumption-free bounds for any contrast between the probabilities of the potential outcome under exposure and non-exposure when the confounders are missing not at random. We assume that the missingness mechanism is outcome-independent. We also report a sensitivity analysis method to complement our bounds.