Research Training Group π³: Parameter identification - Analysis, algorithms, implementations

Description

This project is part of the DFG Research Training Group pi-3 "Parameter Identification - Analysis, Algorithms, Implementations" (Department of Mathematics and Computer Science, University of Bremen). The project considers causal modelling and inference for multivariate time-dependent data. Causal structures are typically represented graphically, most well-known by directed acyclic graphs (DAGs). Key questions relate to the identification of causal effect parameters from availalable observations given an assumed, or previously learned, DAG. In the latter case we speak of causal discovery, for which numerous algorithms are available. However, most of the existing approaches are not suitable for modelling time-dependent data, for instance due to the impossibility to represent feed-back. We have made a proposal for a temporal causal discovery algorithm and the project investigates ist theoretical and statistical behaviour, including application to a real-world cohort, the IDEFICS/I.Family study.

Funding period

Begin:   April 2021
End:   September 2025

Sponsor

  • German Research Foundation

Contact

Prof. Dr. rer. nat. Vanessa Didelez

Project management (national)

Prof Dr. Dr. h.c. Peter Maass, Universität Bremen

Link

Official website of the RTG pi-3

Selected project-related publications

    Presentations at scientific meetings/conferences (invited)

  • Bang CW, Witte J, Foraita R, Didelez V. Improving causal discovery with temporal background knowledge. Seminar at the Section of Biostatistics, University of Copenhagen, 3 October 2023, Copenhagen, Denmark.

    • Presentations at scientific meetings/conferences

  • Bang CW, Witte J, Foraita R, Didelez V. Improving causal discovery for cohort data. 11th Autumn Workshop of the DGEpi (German Society for Epidemiology), GMDS (German Association for Medical Informatics, Biometry and Epidemiology), IBS-DR (German Region of the International Biometric Society) and DGSMP (German Society for Social Medicine and Prevention), 9-10 November 2023, Mainz.
  • Bang CW, Witte J, Foraita R, Didelez V. Improving causal discovery with temporal background knowledge. 5th European Causal Inference Meeting (EuroCIM), 18-21 April 2023, Oslo, Norway.

    • Posters at scientific meetings/conferences

  • Bang CW, Didelez V. Do we become wiser with time? On causal equivalence with tiered background knowledge. 39th Conference on Uncertainty in Artificial Intelligence (UAI), 31 July-4 August 2023, Pittsburgh, USA.
  • Bang CW, Witte J, Foraita R, Didelez V. Efficient use of temporal background knowledge for causal discovery with cohort data. American Causal Inference Conference (ACIC), 23-25 May 2022, Berkeley, USA.