No unobserved single-cause confounders

The 'Substitute' Confounder

1. Assumption 1 : No unobserved single-cause confounders
2. Assumption 2 : The 'substitute' confounder

University of Michigan - Ann Arbor

Towards Clarifying the Theory of the Deconfounder

Wang and Blei previously published a paper that proposed the deconfounder algorithm. This paper builds on that one and clarifies the theory.
It uses the assumption of "no unobserved single-cause confounders" to establish their algorithm's validity.

Key idea of paper

1. The blessings of multiple causes - Wang, Blei 2019 : 
2. The challenges of multiple causes -Imai, Jiang 2019 : 


Important Reference Papers

Multi-Cause Separator is a random variable defined as follows :

Learn Before

Related

Learn After