Confounder Analysis in Measuring Representation in Product Funnels

TL;DR

This paper explores using Shapley values for confounder selection in causal inference, demonstrating their effectiveness in ranking confounders within a scalable matching method using LinkedIn data.

Contribution

It introduces a novel application of Shapley values for confounder importance ranking in coarsened exact matching, enhancing causal inference methods.

Findings

Shapley values effectively identify important confounders.

The method is scalable and applicable to real-world observational data.

Shapley-based ranking improves causal inference robustness.

Abstract

This paper discusses an application of Shapley values in the causal inference field, specifically on how to select the top confounder variables for coarsened exact matching method in a scalable way. We use a dataset from an observational experiment involving LinkedIn members as a use case to test its applicability, and show that Shapley values are highly informational and can be leveraged for its robust importance-ranking capability.