Testable Implications of Causal Graphs in the Presence of Measurement Error

Karthika Mohan, Samuel Zink
Proceedings of the 13th International Conference on Probabilistic Graphical Models, PMLR 343:663-676, 2026.

Abstract

Measurement errors from sensors, instruments and human factors pose significant challenges for causal inference. Classical testable implications such as d-separations require fully observed variables and cannot be readily evaluated when variables are corrupted by measurement error. This paper shows that causal models nevertheless admit testable implications in such settings. We derive novel testable implications for causal graphs and present a general theorem stating sufficient conditions for their existence. Our results enable model falsification when traditional methods are inapplicable, enabling assessment of model compatibility with observed data when measurement error is present.

Cite this Paper

BibTeX
@InProceedings{pmlr-v343-mohan26a, title = {Testable Implications of Causal Graphs in the Presence of Measurement Error}, author = {Mohan, Karthika and Zink, Samuel}, booktitle = {Proceedings of the 13th International Conference on Probabilistic Graphical Models}, pages = {663--676}, year = {2026}, editor = {Camps-Valls, Gustau and Leonelli, Manuele and Varando, Gherardo}, volume = {343}, series = {Proceedings of Machine Learning Research}, month = {09--11 Sep}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v343/main/assets/mohan26a/mohan26a.pdf}, url = {https://proceedings.mlr.press/v343/mohan26a.html}, abstract = {Measurement errors from sensors, instruments and human factors pose significant challenges for causal inference. Classical testable implications such as d-separations require fully observed variables and cannot be readily evaluated when variables are corrupted by measurement error. This paper shows that causal models nevertheless admit testable implications in such settings. We derive novel testable implications for causal graphs and present a general theorem stating sufficient conditions for their existence. Our results enable model falsification when traditional methods are inapplicable, enabling assessment of model compatibility with observed data when measurement error is present.} }
Endnote
%0 Conference Paper %T Testable Implications of Causal Graphs in the Presence of Measurement Error %A Karthika Mohan %A Samuel Zink %B Proceedings of the 13th International Conference on Probabilistic Graphical Models %C Proceedings of Machine Learning Research %D 2026 %E Gustau Camps-Valls %E Manuele Leonelli %E Gherardo Varando %F pmlr-v343-mohan26a %I PMLR %P 663--676 %U https://proceedings.mlr.press/v343/mohan26a.html %V 343 %X Measurement errors from sensors, instruments and human factors pose significant challenges for causal inference. Classical testable implications such as d-separations require fully observed variables and cannot be readily evaluated when variables are corrupted by measurement error. This paper shows that causal models nevertheless admit testable implications in such settings. We derive novel testable implications for causal graphs and present a general theorem stating sufficient conditions for their existence. Our results enable model falsification when traditional methods are inapplicable, enabling assessment of model compatibility with observed data when measurement error is present.
APA
Mohan, K. & Zink, S.. (2026). Testable Implications of Causal Graphs in the Presence of Measurement Error. Proceedings of the 13th International Conference on Probabilistic Graphical Models, in Proceedings of Machine Learning Research 343:663-676 Available from https://proceedings.mlr.press/v343/mohan26a.html.

Related Material