Scalable Causal Structure Learning via Amortized Conditional Independence Testing

James Leiner, Brian Manzo, Aaditya Ramdas, Wesley Tansey
Proceedings of the Fourth Conference on Causal Learning and Reasoning, PMLR 275:174-200, 2025.

Abstract

Controlling false positives (Type I errors) through statistical hypothesis testing is a foundation of modern scientific data analysis. Existing causal structure discovery algorithms either do not provide Type I error control or cannot scale to the size of modern scientific datasets. We consider a variant of the causal discovery problem with two sets of nodes, where the only edges of interest form a bipartite causal subgraph between the sets. We develop Scalable Causal Structure Learning (SCSL), a method for causal structure discovery on bipartite subgraphs that provides Type I error control. SCSL recasts the discovery problem as a simultaneous hypothesis testing problem and uses discrete optimization over the set of possible confounders to obtain an upper bound on the test statistic for each edge. Semi-synthetic simulations demonstrate that SCSL scales to handle graphs with hundreds of nodes while maintaining error control and good power. We demonstrate the practical applicability of the method by applying it to a cancer dataset to reveal connections between somatic gene mutations and metastases to different tissues.

Cite this Paper

BibTeX
@InProceedings{pmlr-v275-leiner25a, title = {Scalable Causal Structure Learning via Amortized Conditional Independence Testing}, author = {Leiner, James and Manzo, Brian and Ramdas, Aaditya and Tansey, Wesley}, booktitle = {Proceedings of the Fourth Conference on Causal Learning and Reasoning}, pages = {174--200}, year = {2025}, editor = {Huang, Biwei and Drton, Mathias}, volume = {275}, series = {Proceedings of Machine Learning Research}, month = {07--09 May}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v275/main/assets/leiner25a/leiner25a.pdf}, url = {https://proceedings.mlr.press/v275/leiner25a.html}, abstract = {Controlling false positives (Type I errors) through statistical hypothesis testing is a foundation of modern scientific data analysis. Existing causal structure discovery algorithms either do not provide Type I error control or cannot scale to the size of modern scientific datasets. We consider a variant of the causal discovery problem with two sets of nodes, where the only edges of interest form a bipartite causal subgraph between the sets. We develop Scalable Causal Structure Learning (SCSL), a method for causal structure discovery on bipartite subgraphs that provides Type I error control. SCSL recasts the discovery problem as a simultaneous hypothesis testing problem and uses discrete optimization over the set of possible confounders to obtain an upper bound on the test statistic for each edge. Semi-synthetic simulations demonstrate that SCSL scales to handle graphs with hundreds of nodes while maintaining error control and good power. We demonstrate the practical applicability of the method by applying it to a cancer dataset to reveal connections between somatic gene mutations and metastases to different tissues.} }
Endnote
%0 Conference Paper %T Scalable Causal Structure Learning via Amortized Conditional Independence Testing %A James Leiner %A Brian Manzo %A Aaditya Ramdas %A Wesley Tansey %B Proceedings of the Fourth Conference on Causal Learning and Reasoning %C Proceedings of Machine Learning Research %D 2025 %E Biwei Huang %E Mathias Drton %F pmlr-v275-leiner25a %I PMLR %P 174--200 %U https://proceedings.mlr.press/v275/leiner25a.html %V 275 %X Controlling false positives (Type I errors) through statistical hypothesis testing is a foundation of modern scientific data analysis. Existing causal structure discovery algorithms either do not provide Type I error control or cannot scale to the size of modern scientific datasets. We consider a variant of the causal discovery problem with two sets of nodes, where the only edges of interest form a bipartite causal subgraph between the sets. We develop Scalable Causal Structure Learning (SCSL), a method for causal structure discovery on bipartite subgraphs that provides Type I error control. SCSL recasts the discovery problem as a simultaneous hypothesis testing problem and uses discrete optimization over the set of possible confounders to obtain an upper bound on the test statistic for each edge. Semi-synthetic simulations demonstrate that SCSL scales to handle graphs with hundreds of nodes while maintaining error control and good power. We demonstrate the practical applicability of the method by applying it to a cancer dataset to reveal connections between somatic gene mutations and metastases to different tissues.
APA
Leiner, J., Manzo, B., Ramdas, A. & Tansey, W.. (2025). Scalable Causal Structure Learning via Amortized Conditional Independence Testing. Proceedings of the Fourth Conference on Causal Learning and Reasoning, in Proceedings of Machine Learning Research 275:174-200 Available from https://proceedings.mlr.press/v275/leiner25a.html.

Related Material