Improved Causal Discovery from Longitudinal Data Using a Mixture of DAGs

Eric V. Strobl
Proceedings of Machine Learning Research, PMLR 104:100-133, 2019.

Abstract

Many causal processes in biomedicine contain cycles and evolve. However, most causal discovery algorithms assume that the underlying causal process follows a single directed acyclic graph (DAG) that does not change over time. The algorithms can therefore infer erroneous causal relations with high confidence when run on real biomedical data. In this paper, I relax the single DAG assumption by modeling causal processes using a mixture of DAGs so that the graph can change over time. I then describe a causal discovery algorithm called Causal Inference over Mixtures (CIM) to infer causal structure from a mixture of DAGs using longitudinal data. CIM improves the accuracy of causal discovery on both real and synthetic clinical datasets even when cycles, non-stationarity, non-linearity, latent variables and selection bias exist simultaneously. Code is available at https://github.com/ericstrobl/CIM.

Cite this Paper

BibTeX
@InProceedings{pmlr-v104-strobl19a, title = {Improved Causal Discovery from Longitudinal Data Using a Mixture of DAGs}, author = {Strobl, Eric V.}, booktitle = {Proceedings of Machine Learning Research}, pages = {100--133}, year = {2019}, editor = {}, volume = {104}, series = {Proceedings of Machine Learning Research}, month = {05 Aug}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v104/strobl19a/strobl19a.pdf}, url = {https://proceedings.mlr.press/v104/strobl19a.html}, abstract = {Many causal processes in biomedicine contain cycles and evolve. However, most causal discovery algorithms assume that the underlying causal process follows a single directed acyclic graph (DAG) that does not change over time. The algorithms can therefore infer erroneous causal relations with high confidence when run on real biomedical data. In this paper, I relax the single DAG assumption by modeling causal processes using a mixture of DAGs so that the graph can change over time. I then describe a causal discovery algorithm called Causal Inference over Mixtures (CIM) to infer causal structure from a mixture of DAGs using longitudinal data. CIM improves the accuracy of causal discovery on both real and synthetic clinical datasets even when cycles, non-stationarity, non-linearity, latent variables and selection bias exist simultaneously. Code is available at https://github.com/ericstrobl/CIM.} }
Endnote
%0 Conference Paper %T Improved Causal Discovery from Longitudinal Data Using a Mixture of DAGs %A Eric V. Strobl %B Proceedings of Machine Learning Research %C Proceedings of Machine Learning Research %D 2019 %E %F pmlr-v104-strobl19a %I PMLR %P 100--133 %U https://proceedings.mlr.press/v104/strobl19a.html %V 104 %X Many causal processes in biomedicine contain cycles and evolve. However, most causal discovery algorithms assume that the underlying causal process follows a single directed acyclic graph (DAG) that does not change over time. The algorithms can therefore infer erroneous causal relations with high confidence when run on real biomedical data. In this paper, I relax the single DAG assumption by modeling causal processes using a mixture of DAGs so that the graph can change over time. I then describe a causal discovery algorithm called Causal Inference over Mixtures (CIM) to infer causal structure from a mixture of DAGs using longitudinal data. CIM improves the accuracy of causal discovery on both real and synthetic clinical datasets even when cycles, non-stationarity, non-linearity, latent variables and selection bias exist simultaneously. Code is available at https://github.com/ericstrobl/CIM.
APA
Strobl, E.V.. (2019). Improved Causal Discovery from Longitudinal Data Using a Mixture of DAGs. Proceedings of Machine Learning Research, in Proceedings of Machine Learning Research 104:100-133 Available from https://proceedings.mlr.press/v104/strobl19a.html.

Related Material