Generalized Instrumental Variables

TL;DR

This paper generalizes the Instrumental Variables method to handle models with limited conditional independences, integrating non-experimental data and qualitative domain knowledge encoded as DAGs, even with unobserved variables.

Contribution

It introduces a generalized IV approach that extends applicability to models with few conditional independences and unobserved variables, guided by domain knowledge.

Findings

Enables causal effect estimation with limited conditional independences.

Incorporates qualitative domain knowledge via DAGs.

Handles models with unobserved variables.

Abstract

This paper concerns the assessment of direct causal effects from a combination of: (i) non-experimental data, and (ii) qualitative domain knowledge. Domain knowledge is encoded in the form of a directed acyclic graph (DAG), in which all interactions are assumed linear, and some variables are presumed to be unobserved. We provide a generalization of the well-known method of Instrumental Variables, which allows its application to models with few conditional independeces.