Deep IV: A Flexible Approach for Counterfactual Prediction

Jason Hartford, Greg Lewis, Kevin Leyton-Brown, Matt Taddy
Proceedings of the 34th International Conference on Machine Learning, PMLR 70:1414-1423, 2017.

Abstract

Counterfactual prediction requires understanding causal relationships between so-called treatment and outcome variables. This paper provides a recipe for augmenting deep learning methods to accurately characterize such relationships in the presence of instrument variables (IVs) – sources of treatment randomization that are conditionally independent from the outcomes. Our IV specification resolves into two prediction tasks that can be solved with deep neural nets: a first-stage network for treatment prediction and a second-stage network whose loss function involves integration over the conditional treatment distribution. This Deep IV framework allows us to take advantage of off-the-shelf supervised learning techniques to estimate causal effects by adapting the loss function. Experiments show that it outperforms existing machine learning approaches.

Cite this Paper


BibTeX
@InProceedings{pmlr-v70-hartford17a, title = {Deep {IV}: A Flexible Approach for Counterfactual Prediction}, author = {Jason Hartford and Greg Lewis and Kevin Leyton-Brown and Matt Taddy}, booktitle = {Proceedings of the 34th International Conference on Machine Learning}, pages = {1414--1423}, year = {2017}, editor = {Precup, Doina and Teh, Yee Whye}, volume = {70}, series = {Proceedings of Machine Learning Research}, month = {06--11 Aug}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v70/hartford17a/hartford17a.pdf}, url = {https://proceedings.mlr.press/v70/hartford17a.html}, abstract = {Counterfactual prediction requires understanding causal relationships between so-called treatment and outcome variables. This paper provides a recipe for augmenting deep learning methods to accurately characterize such relationships in the presence of instrument variables (IVs) – sources of treatment randomization that are conditionally independent from the outcomes. Our IV specification resolves into two prediction tasks that can be solved with deep neural nets: a first-stage network for treatment prediction and a second-stage network whose loss function involves integration over the conditional treatment distribution. This Deep IV framework allows us to take advantage of off-the-shelf supervised learning techniques to estimate causal effects by adapting the loss function. Experiments show that it outperforms existing machine learning approaches.} }
Endnote
%0 Conference Paper %T Deep IV: A Flexible Approach for Counterfactual Prediction %A Jason Hartford %A Greg Lewis %A Kevin Leyton-Brown %A Matt Taddy %B Proceedings of the 34th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2017 %E Doina Precup %E Yee Whye Teh %F pmlr-v70-hartford17a %I PMLR %P 1414--1423 %U https://proceedings.mlr.press/v70/hartford17a.html %V 70 %X Counterfactual prediction requires understanding causal relationships between so-called treatment and outcome variables. This paper provides a recipe for augmenting deep learning methods to accurately characterize such relationships in the presence of instrument variables (IVs) – sources of treatment randomization that are conditionally independent from the outcomes. Our IV specification resolves into two prediction tasks that can be solved with deep neural nets: a first-stage network for treatment prediction and a second-stage network whose loss function involves integration over the conditional treatment distribution. This Deep IV framework allows us to take advantage of off-the-shelf supervised learning techniques to estimate causal effects by adapting the loss function. Experiments show that it outperforms existing machine learning approaches.
APA
Hartford, J., Lewis, G., Leyton-Brown, K. & Taddy, M.. (2017). Deep IV: A Flexible Approach for Counterfactual Prediction. Proceedings of the 34th International Conference on Machine Learning, in Proceedings of Machine Learning Research 70:1414-1423 Available from https://proceedings.mlr.press/v70/hartford17a.html.

Related Material