IMC Researchers win Best Paper Award
Interdisciplinary team of IMC researchers take the prize at COMPLEXIS 2016
The article "Inferring Causality from Noisy Time Series Data" was presented by lead author Dan Mønster at the 1st International Conference on Complex Information Systems — COMPLEXIS 2016, in Rome; and received the best paper award. The article is the product of a collaboration between researchers from four different departments across three faculties, with ties to Interacting Minds Centre.
Detecting causal influence
In the article the authors test Convergent Cross-Mapping, which is a new method to detect causal influence between variables described by time series data. "The method has great potential for detecting causal influence without needing a model of the system under investigation" says lead author Dan Mønster, and continues" "but we wanted to test the method to see how it performs in a model system, where we can control both the direction and strength of the causality between two variables, and we found some important cases where the method breaks down." That the method does not always work is no surprise, since all methods have limitations, and crucially the authors find that inapplicability of the method is signalled by a deviation in the convergence properties of a correlation function. "Having a clear signal that the method does not apply is very valuable, since otherwise we would be left guessing whether the results produced are right or wrong. Now we have found an indicator that warns us when not to trust the results" says Dan Mønster. The authors also show how noise influences the method, and find that correctly inferring causality can—surprisingly—improve in the presence of noise.
Causality — a tricky concept
Even though the article deals with detecting causality, Dan Mønster is reluctant to provide a firm answer to why they were awarded the best paper award. "Causality is a very tricky concept, and so I can only speculate, but the unique research enviroment at Interacting Minds Centre played a very crucial part in our research" he says.
Download and contact information
Dan Mønster, Riccardo Fusaroli, Kristian Tylén, Andreas Roepstorff, and Jacob F. Sherson. (2016). "Inferring causality from noisy time series data." In Proceedings of the 1st International Conference on Complex Information Systems (COMPLEXIS 2016), 48-56.
The article can be downloaded here.
The article is also available as an arXiv preprint: arXiv:1603.01155.
Link to COMPLEXIS 2016 - 1st International Conference on Complex Information Systems
Contact information:
- Assistant Professor Dan Mønster, Dept. of Economics and Business Economics, IMC and Cognition and Behavior Lab
Co-authors
- Assistant Professor Riccardo Fusaroli, School of Culture and Communikation
- Associate Professor Kristian Tylén, School of Communication and Culture
- Professor Andreas Roepstorff, Dept. of Clinical Medicine and School of Culture and Society
- Associate Professor Jacob Sherson, Dept. of Physics and Astronomy
