Identification In Missing Data Models Represented By Directed Acyclic Graphs

Rohit Bhattacharya, Razieh Nabi, Ilya Shpitser, James M. Robins
Proceedings of The 35th Uncertainty in Artificial Intelligence Conference, PMLR 115:1149-1158, 2020.

Abstract

Missing data is a pervasive problem in data analyses, resulting in datasets that contain censored realizations of a target distribution. Many approaches to inference on the target distribution using censored observed data, rely on missing data models represented as a factorization with respect to a directed acyclic graph. In this paper we consider the identifiability of the target distribution within this class of models, and show that the most general identification strategies proposed so far retain a significant gap in that they fail to identify a wide class of identifiable distributions. To address this gap, we propose a new algorithm that significantly generalizes the types of manipulations used in the ID algorithm [14, 16], developed in the context of causal inference, in order to obtain identification.

Cite this Paper

BibTeX
@InProceedings{pmlr-v115-bhattacharya20b, title = {Identification In Missing Data Models Represented By Directed Acyclic Graphs}, author = {Bhattacharya, Rohit and Nabi, Razieh and Shpitser, Ilya and Robins, James M.}, booktitle = {Proceedings of The 35th Uncertainty in Artificial Intelligence Conference}, pages = {1149--1158}, year = {2020}, editor = {Adams, Ryan P. and Gogate, Vibhav}, volume = {115}, series = {Proceedings of Machine Learning Research}, month = {22--25 Jul}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v115/bhattacharya20b/bhattacharya20b.pdf}, url = {https://proceedings.mlr.press/v115/bhattacharya20b.html}, abstract = {Missing data is a pervasive problem in data analyses, resulting in datasets that contain censored realizations of a target distribution. Many approaches to inference on the target distribution using censored observed data, rely on missing data models represented as a factorization with respect to a directed acyclic graph. In this paper we consider the identifiability of the target distribution within this class of models, and show that the most general identification strategies proposed so far retain a significant gap in that they fail to identify a wide class of identifiable distributions. To address this gap, we propose a new algorithm that significantly generalizes the types of manipulations used in the ID algorithm [14, 16], developed in the context of causal inference, in order to obtain identification.} }
Endnote
%0 Conference Paper %T Identification In Missing Data Models Represented By Directed Acyclic Graphs %A Rohit Bhattacharya %A Razieh Nabi %A Ilya Shpitser %A James M. Robins %B Proceedings of The 35th Uncertainty in Artificial Intelligence Conference %C Proceedings of Machine Learning Research %D 2020 %E Ryan P. Adams %E Vibhav Gogate %F pmlr-v115-bhattacharya20b %I PMLR %P 1149--1158 %U https://proceedings.mlr.press/v115/bhattacharya20b.html %V 115 %X Missing data is a pervasive problem in data analyses, resulting in datasets that contain censored realizations of a target distribution. Many approaches to inference on the target distribution using censored observed data, rely on missing data models represented as a factorization with respect to a directed acyclic graph. In this paper we consider the identifiability of the target distribution within this class of models, and show that the most general identification strategies proposed so far retain a significant gap in that they fail to identify a wide class of identifiable distributions. To address this gap, we propose a new algorithm that significantly generalizes the types of manipulations used in the ID algorithm [14, 16], developed in the context of causal inference, in order to obtain identification.
APA
Bhattacharya, R., Nabi, R., Shpitser, I. & Robins, J.M.. (2020). Identification In Missing Data Models Represented By Directed Acyclic Graphs. Proceedings of The 35th Uncertainty in Artificial Intelligence Conference, in Proceedings of Machine Learning Research 115:1149-1158 Available from https://proceedings.mlr.press/v115/bhattacharya20b.html.

Related Material